Photo Acoustic and Optical

Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
 Photo Acoustic and Optical
Coherence Tomography Imaging,
Volume 3
Angiography: an application in vessel imaging

Edited by
Ayman El-Baz
University of Louisville, Louisville, KY, USA
and
University of Louisville at AlAlamein International University (UofL-AIU),
New Alamein City, Egypt

Jasjit S Suri
AtheroPoint LLC, Roseville, CA, USA

IOP Publishing, Bristol, UK

ª IOP Publishing Ltd 2021

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system
or transmitted in any form or by any means, electronic, mechanical, photocopying, recording
or otherwise, without the prior permission of the publisher, or as expressly permitted by law or
under terms agreed with the appropriate rights organization. Multiple copying is permitted in
accordance with the terms of licences issued by the Copyright Licensing Agency, the Copyright
Clearance Centre and other reproduction rights organizations.

Permission to make use of IOP Publishing content other than as set out above may be sought
at [email protected].

Ayman El-Baz and Jasjit S Suri have asserted their right to be identified as the editors of this work
in accordance with sections 77 and 78 of the Copyright, Designs and Patents Act 1988.

ISBN 978-0-7503-2060-3 (ebook)

ISBN 978-0-7503-2058-0 (print)
ISBN 978-0-7503-2059-7 (myPrint)
ISBN 978-0-7503-2061-0 (mobi)

DOI 10.1088/978-0-7503-2060-3

Version: 20211201

IOP ebooks

British Library Cataloguing-in-Publication Data: A catalogue record for this book is available
from the British Library.

Published by IOP Publishing, wholly owned by The Institute of Physics, London

IOP Publishing, Temple Circus, Temple Way, Bristol, BS1 6HG, UK

US Office: IOP Publishing, Inc., 190 North Independence Mall West, Suite 601, Philadelphia,
PA 19106, USA
With love and affection to my mother and father, whose loving spirit sustains me still
— Ayman El-Baz

To my late loving parents, immediate family, and children

— Jasjit S Suri
 Contents

Preface xiv
Acknowledgements xv
Editor biographies xvi
List of contributors xvii

1 Clinical application of optical coherence tomography 1-1

angiography in retinal diseases
Hashim Ali Khan, Muhammad Aamir Shahzad, Muhammad Amer Awan
and Smaha Jahangir
1.1 Introduction 1-1
1.2 Comparison of OCTA with FA and ICGA 1-2
1.3 Principles and methods of OCTA 1-2
1.3.1 Phase based strategies 1-3
1.3.2 Intensity/amplitude based strategies 1-4
1.3.3 Complex signal-based OCTA 1-4
1.4 Commercial devices 1-4
1.4.1 RTVue XR Avanti 1-4
1.4.2 Topcon Triton SS OCTAngio™ 1-5
1.4.3 ZEISS AngioPlex™ 1-5
1.4.4 Heidelberg Spectralis® OCTA 1-5
1.4.5 Nidek angioscan 1-5
1.5 OCTA of normal retina 1-5
1.5.1 Retinal and choroidal vasculature 1-5
1.5.2 Default OCTA segmentations 1-6
1.5.3 Artifacts 1-6
1.6 OCTA in retinal vascular diseases 1-9
1.6.1 Diabetic retinopathy 1-10
1.6.2 Retinal vein occlusion (RVO) 1-10
1.6.3 Retinal artery occlusion 1-11
1.6.4 Central serous chorioretinopathy 1-12
1.6.5 Macular telangiectasia 1-13
1.6.6 Other retinal diseases 1-13
1.7 Choroidal diseases 1-14
1.7.1 Dry age-related macular degeneration and variants 1-14
1.7.2 Wet AMD and variants 1-16

vii
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

1.8 OCTA in ocular inflammation 1-19

1.9 Conclusion 1-20
References 1-20

2 Longitudinal optical coherence tomography and angiography 2-1

of hyaloid vascular regression in developing mouse eyes
Tae-Hoon Kim and Xincheng Yao
2.1 Introduction 2-2
2.2 Methods 2-3
2.2.1 OCT/OCTA imaging system 2-3
2.2.2 OCT/OCTA image acquisition 2-4
2.2.3 Quantitative image analysis 2-4
2.2.4 Animals 2-6
2.3 Results 2-6
2.3.1 OCTA monitoring of hyaloid vascular regression 2-6
2.3.2 Accelerated regression of hyaloid vessels in retinal 2-8
degeneration mice
2.4 Discussion 2-10
2.5 Conclusion 2-12
References 2-12

3 Stripe noise removal and vessel segmentation of 3-1

OCTA images
Xiyin Wu, Dongxu Gao, Davide Borroni, Savita Madhusudhan
and Yalin Zheng
3.1 Introduction 3-1
3.2 Two-stage strategy (TSS) 3-4
3.2.1 The first stage: stripe noise removal of OCTA images 3-5
3.2.2 The second stage: segmentation of vessels from stripe denoised 3-8
images
3.3 Joint destriping and segmentation of OCTA images (JDS) 3-10
3.4 Results 3-14
3.4.1 Datasets 3-14
3.4.2 Effectiveness of destriping 3-15
3.4.3 Effectiveness of segmentation 3-18
3.5 Discussion and conclusions 3-22
References 3-23

viii
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

4 Optical coherence tomography angiography changes in 4-1

early type 3 neovascularization after anti-vascular
endothelial growth factor treatment
Riccardo Sacconi, Carlo Di Biase, Enrico Borrelli, Lea Querques, Francesco
Prascina, Ilaria Zucchiatt, Francesco Bandello and Giuseppe Querques
4.1 Introduction 4-2
4.2 Nascent type 3 MNV 4-2
4.3 Treatment of active type 3 MNV 4-4
4.4 OCT-A features of recurrent type 3 MNV after anti-VEGF injections 4-5
References 4-8

5 Optical coherence tomography angiography in multiple 5-1

sclerosis and neuromyelitis optica
Olwen C Murphy and Shiv Saidha
5.1 Introduction 5-1
5.2 Optical coherence tomography (OCT) of the retina 5-3
5.2.1 Retinal OCT in multiple sclerosis (MS) 5-4
5.2.2 Retinal OCT in neuromyelitis optica spectrum 5-4
disorder (NMOSD)
5.3 Optical coherence tomography angiography (OCTA) of the retina 5-5
5.3.1 OCTA in MS 5-6
5.3.2 OCTA in NMOSD 5-10
5.4 Limitations of OCTA 5-10
5.5 Conclusions 5-12
References 5-12

6 Optical coherence tomography angiography for the 6-1

diagnosis of polypoidal choroidal vasculopathy
Talisa De Carlo and Gregg T Kokame
6.1 Introduction 6-1
6.2 OCTA for the diagnosis of PCV 6-2
6.2.1 OCT angiograms 6-2
6.2.2 OCT b-scans 6-4
6.2.3 OCTA image set 6-4
6.3 Discussion 6-6
6.4 Conclusion 6-6
References 6-6

ix
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

7 Quantitative features for objective assessment of OCT 7-1

angiography
Minhaj Nur Alam, David Le, Taeyoon Son, Jennifer I Lim and
Xincheng Yao
7.1 Introduction 7-1
7.2 Quantitative OCTA features 7-2
7.2.1 Blood vessel density 7-3
7.2.2 Blood vessel caliber 7-3
7.2.3 Blood vessel tortuosity 7-4
7.2.4 Vessel perimeter index 7-4
7.2.5 Foveal avascular zone features 7-5
7.2.6 Vessel complexity features 7-5
7.2.7 Branchpoint analysis 7-6
7.2.8 Flow analysis 7-7
7.2.9 Choroidal vasculature 7-8
7.2.10 Differential artery–vein (A–V) analysis 7-8
7.3 Machine learning techniques in objective OCTA analysis 7-10
7.4 Discussion 7-13
References 7-15

8 Clinical utility of OCT-angio in age-related 8-1

macular degeneration
Hashim Ali Khan, Smaha Jahangir and Julie Friedman Rodman
8.1 Clinical features and OCTA correlates of early AMD 8-2
8.1.1 Role of chorioretinal flow characteristic in pathogenesis 8-4
of drusen
8.2 Geographic atrophy 8-5
8.3 Neovascular AMD 8-6
8.3.1 Type 1 MNV 8-7
8.3.2 Type 2 MNV 8-9
8.3.3 Type 3 MNV 8-10
8.4 Diagnostic accuracy of OCTA 8-11
8.5 OCTA for guiding treatment of AMD 8-13
References 8-13

x
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

9 Optical coherence tomography angiography imaging in 9-1

age-related macular degeneration
Onur Gokmen and Muhammet Derda Ozer
9.1 Introduction 9-1
9.2 The principles Of OCTA technology and image quantification methods 9-1
9.3 FFA, ICG, and OCTA 9-2
9.4 AMD types 9-3
9.5 OCTA features in AMD 9-3
9.6 Dry-type age-related macular degeneration 9-4
9.7 Wet-type age-related macular degeneration 9-5
9.8 Conclusion 9-7
References 9-8

10 Optical coherence tomography angiography in 10-1

ophthalmology: an application in vessel imaging
Anadi Khatri, Araniko Pandey, Gunjan Prasai, Kinsuk Singh,
Muna Kharel, Eli Pradhan and Rupesh Agrawal
10.1 Introduction 10-1
10.2 Principles of OCTA 10-3
10.2.1 Variants 10-5
10.2.2 Fundamentals of OCTA in ophthalmology 10-5
10.2.3 Scanning technique 10-6
10.3 OCTA of normal eyes 10-9
10.4 Current applications of OCTA in ophthalmology 10-9
10.4.1 OCTA in dry (non-neovascular) AMD 10-9
10.4.2 OCTA in wet (neovascular) AMD 10-10
10.4.3 OCTA in diabetic retinopathy 10-11
10.4.4 OCTA in artery and vein occlusion 10-12
10.4.5 OCTA in glaucoma 10-13
10.4.6 OCTA in uveitis 10-14
10.5 The future of OCTA and ophthalmology 10-16
10.5.1 Non-invasive imaging and neurovascular coupling 10-16
References 10-17

11 Optical coherence tomography angiography: principles and 11-1

clinical application
Ogugua Ndubuisi Okonkwo and Adekunle Olubola Hassan
11.1 Introduction 11-2

xi
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

11.2 Principles of OCTA 11-2

11.2.1 Basics 11-2
11.2.2 Image segmentation 11-3
11.2.3 En face imaging 11-4
11.3 OCTA technology: software and hardware 11-5
11.4 FA versus OCTA comparison 11-6
11.4.1 Limitations of OCTA 11-9
11.4.2 A summary of OCTA artifacts 11-10
11.5 OCTA imaging of the normal retina and optic disc 11-10
11.6 OCTA clinical application 11-12
11.6.1 Uveitis 11-12
11.6.2 Diabetic retinopathy 11-12
11.6.3 Macular degeneration 11-13
11.6.4 Retinal vein occlusion 11-15
11.6.5 Retinal artery occlusion 11-22
11.6.6 Vitreomacular interface disease 11-22
11.6.7 Glaucomatous optic neuropathy (GON) 11-26
11.6.8 Miscellaneous 11-27
11.7 Discussion 11-27
11.8 Conclusion 11-30
References 11-30

12 A comprehensive survey on computer-aided diagnostic 12-1

systems in diabetic retinopathy screening
Meysam Tavakoli and Patrick Kelley
12.1 Introduction 12-2
12.1.1 Clinical signs of diabetic retinopathy 12-3
12.1.2 Stages of diabetic retinopathy 12-4
12.1.3 Retinal imaging and databases 12-6
12.2 Computer-aided diagnosis and diabetic retinopathy 12-9
12.3 Retinal landmarks detection 12-13
12.3.1 Localization and segmentation of optic nerve head 12-13
12.3.2 Vessel segmentation 12-16
12.3.3 Localization of the fovea 12-20
12.4 Detection of retinal lesions 12-21
12.4.1 Microaneurysm and haemorrhage detection 12-21
12.4.2 Exudates segmentation 12-24

xii
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

12.5 Artificial intelligence in diabetic retinopathy 12-26

12.5.1 Overview of deep learning 12-26
12.6 Conclusion 12-31
References 12-31

xiii
 Preface

The book covers the state-of-the-art techniques of optical coherence tomography

angiography (OCTA) imaging for the diagnosis of retinal diseases. Clinical disorders
of the retina have been attracting the attention of researchers, aiming at reducing the
blindness rate. This includes uveitis, diabetic retinopathy, macular edema, endoph-
thalmitis, proliferative retinopathy, age-related macular degeneration and glau-
coma. Currently, most ophthalmologists perform diagnosis by visual observation
and interpretation. Treatment is significantly dependent on having early and
accurate diagnosis, which can be significantly improved by employing disease-
specific computer-aided diagnostic (CAD) systems based on different image modal-
ities such as: optical coherence tomography (OCT), Fundus Imaging and optical
coherence tomography angiography (OCTA). This book will focus on OCTA
imaging for the diagnosis of any retinal disease. Among the topics discussed in
the book are clinical application of OCTA; hyaloid vascular regression in develop-
ing mouse eyes; OCTA stripe noise removal and vessel segmentation; OCTA
changes in early type 3 neovascularization after anti-vascular endothelial growth
factor treatment; OCTA in multiple sclerosis and neuromyelitis optica; diagnosis of
polypoidal choroidal vasculopathy; objective assessment of OCTA; and age-related
macular degeneration.
In summary, the main aim of this book is to help advance scientific research
within the broad field of OCTA imaging for the diagnosis of retinal diseases. The
book focuses on major trends and challenges in this area, and it presents work aimed
to identify new techniques and their use in biomedical analysis.
Ayman El-Baz
Jasjit S Suri

xiv
 Acknowledgements

The completion of this book could not have been possible without the participation
and assistance of so many people whose names may not all be enumerated. Their
contributions are sincerely appreciated and gratefully acknowledged. However, the
editors would like to express their deep appreciation and indebtedness particularly to
Dr Ali H Mahmoud and Reem Haweel for their endless support.
Ayman El-Baz
Jasjit S Suri

xv
 Editor biographies

Ayman El-Baz
Ayman El-Baz is a Distinguished Professor at University of
Louisville, Kentucky, United States and University of Louisville
at AlAlamein International University (UofL-AIU), New
Alamein City, Egypt. Dr El-Baz earned his BSc and MSc degrees
in electrical engineering in 1997 and 2001, respectively. He earned
his PhD in electrical engineering from the University of Louisville
in 2006. Dr El-Baz was named as a Fellow for Coulter, AIMBE
and NAI for his contributions to the field of biomedical translational research.
Dr El-Baz has almost two decades of hands-on experience in the fields of bio-
imaging modeling and non-invasive computer-assisted diagnosis systems. He has
authored or coauthored more than 500 technical articles (155 journals, 44 books, 85
book chapters, 255 refereed-conference papers, 196 abstracts, and 36 US patents and
Disclosures).

Jasjit S Suri
Jasjit S Suri, PhD, MBA is a Fellow of IEEE, AIMBE, SVM,
AIUM, and APVS. He is currently the Chairman of AtheroPoint,
Roseville, CA, USA, dedicated to imaging technologies for
cardiovascular and stroke. He has nearly ~22,000 citations, has
co-authored 50 books, and has an H-index of 72.

xvi
 List of contributors

Rupesh Agrawal
Tang Tock Seng Hospital, Singapore

Dr Rupesh Agrawal, is practising as Senior Consultant

Ophthalmologist at National Healthcare Group Eye Institute,
Tan Tock Seng Hospital, Singapore. After completing his medical
studies from Nagpur (India), he did his post-graduation and
fellowship in Uveitis and Ocular trauma from Sankara Netralaya,
Chennai (India). He was subsequently working as Consultant Ophthalmologist
(Uveitis, Ocular Trauma and Cataract) at Shri Ganapati Netralaya, Jalna (India).
He also worked briefly at L V Prasad Eye Institute before migrating to Singapore in
2009. Subsequently, he has been working in Singapore and involved in significant
translational and basic science research projects. He completed his overseas research
training fellowship—awarded by National Medical Research Council Scholarship,
Ministry of Health, Singapore at University College London on retinal imaging
using retinal leukogram and erythrogram projects. He is also working on mechan-
ical properties of the red blood cell and its implications in microvascular disorders.
He has to his credit 200+ peer reviewed publications and 50+ book chapters. He
is actively involved in numerous collaborative projects on uveitis. He is recipient of
many grants for numerous basic science projects pertinent to ocular inflammation.
He is currently secretary of International Society of Ocular trauma and Asia Pacific
Ophthalmic Trauma Society.

Minhaj Nur Alam

Department of Biomedical Data Science, Stanford University,
Stanford, CA, USA

Minhaj Nur Alam, PhD is currently a postdoctoral fellow at the

Stanford University, Department of Biomedical Data Science
(Center for Artificial Intelligence in Medical Imaging, AIMI).
He is an alumnus of the Biomedical Optics and Functional
Imaging lab at the University of Illinois at Chicago. His research focus is in the
field of biomedical and optical image processing and medical AI applications. He
has extensive experience in incorporating quantitative image processing and
machine learning algorithms for computer aided diagnosis in ophthalmology/
radiology.

xvii
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Muhammad Amer Awan

Shifa International Hospital, Islamabad, Pakistan

Muhammad Amer Awan, FRCS, FRCOphth, is a Professor of

Ophthalmology and a consultative Vitreoretinal surgeon cur-
rently affiliated with Shifa College of Medicine/Shifa
International Hospital. He is an avid researcher and has pub-
lished extensively in peer-reviewed journals. He has a special
interest in a 27 gauge micro-incision vitrectomy system to manage various retinal
diseases.

Francesco Bandello
Department of Ophthalmology, University Vita-Salute, IRCCS
Ospedale San Raffaele, Milan, Italy

Francesco Bandello, MD, FEBO, is Full Professor and Chairman at

the Department of Ophthalmology University Vita-Salute, Scientific
Institute San Raffaele, Milan, Italy. He is Academic Dean of ‘Corso
di Laurea Specialistica/Magistrale in Medicina e Chirurgia’ at the
same University. Professor Bandello is Past President of EURETINA, President of
Academia Ophthalmologica Europea and Vice-President of EuroLam. Professor
Bandello is Editor-in-Chief of the European Journal of Ophthalmology and former
board member of the Club Jules Gonin and Macula Society. He is member of
Executive Board of ESASO Foundation (European School for Advanced Studies in
Ophthalmology), member of the Academia Ophthalmologica Internationalis and the
Accademia Nazionale di Medicina.
Professor Bandello is co-author of 11 books and he serves as a peer reviewer for
grant applications for the NEI. He has authored or co-authored 500 PubMed
articles and he served as trained Principal Investigator in several clinical trials
performed following ICH/GCP and mainly concerning retinal diseases.

Enrico Borrelli
Department of Ophthalmology, University Vita-Salute, IRCCS
Ospedale San Raffaele, Milan, Italy

Enrico Borrelli, MD, FEBO, graduated from Vita-Salute San

Raffaele University (Milan, Italy) in 2013. He completed his
residency in ophthalmology at Gabriele D’Annunzio University
(Chieti, Italy) and his clinical research fellowship in medical retina
and retinal imaging at Doheny Eye Institute, David Geffen School of Medicine at
UCLA (Los Angeles, CA, USA). Currently, he works as retina specialist at San
Raffaele Hospital (Milan, Italy). His research interests include age-related macular
degeneration and image analysis studies.

xviii
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Davide Borroni
Eye and Vision Science Department, University of Liverpool,
Liverpool, UK

Mr Davide Borroni is an ophthalmologist and researcher in The

Venice Eye Bank Foundation, Venice, Italy. He completed
his Fellowship in Cornea, Cataract and Refractive Surgery in
St Paul’s Eye Unit, Royal Liverpool University Hospital,
Liverpool, UK. He received the 2018 ESCRS Clinical Research Awards for his
studies in metagenomics in ophthalmology. He has experience in corneal transplants
and microbiology.

Talisa de Carlo
University of Colorado at Denver, Denver, CO, USA

Dr de Carlo earned her Bachelors of Arts Degree from the University

of Pennsylvania, where she graduated with Cum Laude honors. She
received her M.D. from Tufts University School of Medicine,
graduating with research honors and awarded the William
Dameshek Undergraduate Research Award. She completed the
Optical Coherence Tomography Research Fellowship at the New England Eye
Center and Massachusetts Institute of Technology during a gap year in medical
school. Dr de Carlo completed her Internship at the University of Hawaii and her
Ophthalmology Residency at the Illinois Eye and Ear Infirmary of the University of
Illinois at Chicago. She is currently a Vitreoretinal Surgery Fellow at the University of
Colorado at Denver. She has authored over 50 manuscripts and presented her
research at 15 international and local ophthalmology meetings. She is also one of
the founding members of the International Ocular Circulation Society and an invited
reviewer for multiple esteemed ophthalmology journals. Some of her accolades
include the University of Illinois at Chicago Hospital Resident of the Year (August
2019), VitreoRetinal Surgery Foundation Research Award (November 2017), and the
Retina Research Foundation/Joseph M and Eula C Lawrence Travel Grant of the
Association for Research in Vision and Ophthalmology (February 2015).

Carlo Di Biase
Department of Ophthalmology, University Vita-Salute, IRCCS
Ospedale San Raffaele, Milan, Italy

Carlo Di Biase, MD, received his Medical degree at University of

Genoa, Italy in 2017, and is currently resident, in the second year,
of ophthalmology at Vita-Salute San Raffaele University in
Milan, Italy.

xix
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Dongxu Gao
Eye and Vision Science Department, University of Liverpool,
Liverpool, UK

Dongxu Gao is a Postdoctoral researcher at the University of

Liverpool, UK. He received his PhD degree in computer science
from the University of Portsmouth in 2017. He has rich experience
in face recognition, object tracking and human motion recogni-
tion. His research interests are deep learning, computer vision and their applications.

Onur Gökmen
Yuzuncu Yil University Faculty of Medicine, Ophthalmology
Department, Van, Turkey

Dr Onur Gökmen studied medicine in Gazi University Faculty of

Medicine, Ankara, Turkey. After completion of his MD degree, he
enrolled in the ophthalmology residency program at Baskent
University Faculty of medicine ophthalmology department. He
fulfiled the government duty at Health Sciences University Van Education and
Research Hospital and worked five years. Later, he worked as a doctor lecturer for
two years at the university of Health Sciences Samsun Training And Research Hospital.
He was an observer at the University of Miami Bascom Palmer Eye Institute and
Cleveland Clinic Cole Eye İnstitute. He has acquired the postnominal acronym FEBO
in 2016. He is married and has two daughters. He has published a dozen scientific papers
in well-known peer-reviewed journals. His specific area of interest is ophthalmic plastic
surgery and medical-surgical retina. He is currently working as an academic staff
member in Van Yuzuncu Yil University, Faculty of Medicine.

Adekunle Hassan
Eye Foundation Retina Institute and Eye Foundation Hospital,
Lagos, Nigeria

Dr Kunle Hassan is an internationally acclaimed ophthalmic and

vitreoretinal surgeon with vast experience in the care of vitreor-
etinal diseases. He has trained several generations of ophthalmic
and retinal surgeons and is the founder and chairman of the first
ophthalmic group practice, with multiple branches across Nigeria. He has several
firsts to his credit, including acquiring the first OCT Angiography technology in the
country and the first indigenous state-of-the-art vitreoretinal surgical facility in
the region. This center receives vitreoretinal referrals from within the country and
the region. Dr Hassan’s passion for eye care providers’ training is evident as he
established the first private ophthalmic residency training program in the country
and established training for other cadres of eye care providers.

xx
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Smaha Jahangir
The University of Faisalabad, Faisalabad, Punjab, Pakistan

Smaha Jahangir, OD, is a dedicated optometric physician and is

specializing in retinal imaging and diseases. She is a post-graduate
student at the University of Faisalabad. She is an astute clinician
with excellent knowledge and command on ocular imaging. Her
research interests are focused on clinical applications of OCT.

Patrick Kelley
Department of Physics, Indiana University-Purdue University,
Indianapolis, IN, USA

Patrick Kelley received his BS degree with a dual major in physics

and mathematics from Purdue University in 2016; he received his
MS degree in educational physics from Purdue University in 2019
performing textual analysis, primarily sentiment analysis, on data
accumulated from a social media platform called CourseNetworking (CN) tailored
to student interaction. He applied his knowledge of text mining toward machine
learning for image processing. He has an active interest in machine learning, deep
learning and its future applications. He is currently working on his PhD in
experimental physics. The project is on optical cooling of a nanoparticle, imple-
menting machine learning, with the goal to use in ultra-weak gravity force
measurements. Of the many skills, he has gathered a strong background in
LabView, Python, R, SAS, MatLab, and Mathematica, with some familiarity
with HTML, Javascript, CSS, and MarkDown.

Hashim Ali Khan

SEHHAT Foundation Hospital, Gilgit, Pakistan

Hashim Ali Khan, OD, FAAO, is a consultative optometrist,

specializing in retinal diseases and retinal imaging. He is an
Assistant Professor of Optometry at the University of
Faisalabad. With a remarkable background in publishing scien-
tific articles, he serves on the editorial boards of the Journal of
Ocular Infection and Inflammation and Journal Inside Optometry. He is a certified
clinical OCT/OCTA grader and provides retinal image reading services to many
centers.

xxi
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Muna Kharel
Nepalese Army Institute of Health Sciences, Kathmandu, Nepal

Dr Muna completed her MBBS from Kathmandu medical college

and teaching hospital, Kathmandu Nepal and is currently con-
sultant ophthalmologist at Birat Eye Hospital. She has a keen
interest in glaucoma and medical retina and has multiple publica-
tions to her name.

Anadi Khatri
Birat Eye Hospital, Biratnagar, Nepal

Dr Anadi Khatri is the Medical Director, co-founder and head of

Vitreo-Retina Services at Birat Eye Hospital, Nepal. He is also
Associate Professor of Ophthalmology at Birat Medical College.
He is also on the editorial boards of Nepalese Journal of
Ophthalmology, BMC ophthalmology and Birat Journal of Health
Sciences. His interests and field of research are in lasers in ophthalmology, multi-
modal imaging, Anti-VEGF agents, retinal vasculopathies and innovations in
vitreo-retina surgery. He has over 45 publications in reputed journals on the same
fields of ophthalmology.
He has multiple awards including American academy of ophthalmology
Eyewiiki international ophthalmologist contest winner, ARVO DCERF, Japanese
Ophthalmological society young investigator award and SAARC Academy of
Ophthalmology Best Research Paper to his name.

Taehoon Kim
Department of Biomedical Engineering, University of Illinois at
Chicago (UIC), Chicago, IL, USA

Taehoon Kim is a PhD candidate at University of Illinois at

Chicago (UIC). Taehoon obtained his BS and MS degrees in
Biomedical Engineering at Catholic University of Daegu. After
graduation, he worked in Texas Tech University as a research
scientist for a year. Taehoon then joined the Yao lab at UIC in August 2017 for his
doctoral study. His current research focuses on optical imaging of retinal function.

xxii
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Gregg Kokame
Retina Consultants of Hawaii; University of Hawaii John A Burns
School of Medicine, Aiea, HI, USA

Dr Kokame earned his Bachelor of Arts Degree from Pomona

College, where he graduated with Phi Beta Kappa and Magna Cum
Laude honors. He received his MD from the UCLA School of
Medicine, where he graduated with honors in the honor medical
society, Alpha Omega Alpha. Dr Kokame completed his ophthalmology residency
at the Jules Stein Eye Institute at the UCLA School of Medicine, consistently rated
as the top eye hospital in the western USA. He completed his Vitreoretinal Surgery
Fellowship at Bascom Palmer Eye Institute at the University of Miami School of
Medicine in Miami, Florida, which was recently rated as the top eye surgeon
hospital and training program in ophthalmology in the USA. Dr Kokame is Chief of
Ophthalmology and Clinical Professor at the University of Hawaii John A Burns
School of Medicine, where he teaches international vitreoretinal fellows, residents,
interns and medical students about retinal diseases. He has directed The Retina
Center at Pali Momi into a nationally and internationally recognized center for
retinal research and education. Many new clinical trials bringing new advances over
the last decade have initially started in Hawaii as clinical trials at The Retina Center
at Pali Momi. He has authored over 100 scientific papers in peer-reviewed journals,
and is a consultant and reviewer for all the major journals in ophthalmology.
Dr Kokame is a member of the American Ophthalmological Society, Association
of University Professors of Ophthalmology, Macula Society, American Academy of
Ophthalmology, and American Society of Retina Specialists. He has received the
Honor and Senior Honor Award of the American Society of Retinal Specialists, and
the Secretariat and Achievement Award and Senior Achievement Award from the
American Academy of Ophthalmology. He has given over 70 invited lectures,
including the Pettit Lecture at the UCLA Jules Stein Eye Institute Annual Seminar
in 2013.

David Le
Department of Biomedical Engineering, University of Illinois at
Chicago, Chicago, IL, USA

David Le is a PhD candidate at the University of Illinois at

Chicago (UIC). David obtained his BS in Biomedical
Engineering from Rutgers University. After graduation, he did
research at Rutgers University related to artificial intelligence in
biomedical imaging. David then joined the Yao lab at UIC in
August 2018 for his doctoral study. His current research focuses on disease
classification using retinal imaging.

xxiii
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Jennifer I Lim
Department of Ophthalmology and Visual Sciences, University of
Illinois at Chicago, Chicago, IL, USA

Dr Jennifer I Lim holds the Marion H Schenk Esq., Chair in

Ophthalmology for Research in the Aging Eye as a Professor of
Ophthalmology, Vice-Chair of Ophthalmology for Diversity &
Inclusion, and Director of the Retina Service at University of
Illinois at Chicago. Her areas of expertise include surgical and medical retinal
diseases. She received her MD with Distinction from Northwestern University
(Six-Years Honors Program in Medical Education), where she was elected into
AOA and received both the Dean’s AOA Student Research Award and the Julius
Conn Memorial Award. She completed her ophthalmology training at University of
Illinois at Chicago (UIC), Illlinois Eye and Ear Infirmary where she received the
Resident Research Award and the American College of Surgeons Resident
Competition Chicago Chapter Keeshin Prize. She completed medical and surgical
retina fellowships at the Wilmer Eye Institute of Johns Hopkins Hospital, where she
was a Heed Fellow and a Heed Knapp Fellow Awardee. After completing her
training, she joined the faculty of Emory University as an Assistant Professor of
Ophthalmology. She was recruited to Doheny Eye Institute of University of
Southern California as an Associate Professor with tenure and then promoted to
Professor of Ophthalmology. In 2007, she was recruited to University of Illinois at
Chicago as the Director of the Retina Service. Her research interests include clinical
trials, translational research, retinal imaging for which she has an R01 NIH grant
(Co-PI on ‘Differential artery-vein analysis in OCT angiography for objective
classification of diabetic retinopathy’, and retinal vascular disease. She has been
principal investigator in over 60 clinical trials, serves on Data and Monitoring
Committees and sits on Executive Committees. She collaborates with basic scientists
on angiogenesis and imaging research. She has received funding from the National
Eye Institute, Macula Society Research Fund, American Cancer Society and
numerous industry grants. Current leadership positions include Associate Editor
for JAMA Ophthalmology, EyeWiki Deputy Editor-in-Chief, Vice-President of The
Retina Society, President of Chicago Ophthalmological Society, IOVS Editorial
Board member, AOA UIC Medical School Councilor, and UIC Faculty Senator.
She is a past-President of both Women in Ophthalmology and Chinese American
Ophthalmology Society and past Chair of Women in Retina.
She has received the AAO Life Achievement Honor Award, AAO Senior
Achievement Award, AAO Secretariat Award, American Society of Retinal
Specialists Senior Honor Award, Suzanne-Veronneau Troutman Award, ARVO
Gold and Silver Fellow Awards, the Paul Henkind Memorial Award, Women in
Ophthalmology Scientific Contribution Award, Chinese American Ophthalmology
Service awards, USC teaching awards, ASRS Retina Hall of Fame, UIC
Departmental Faculty of the Year Award, Mother McAuley Liberal Arts HS
Hall of Honor Award, and honored on the inaugural 2021 Newsweek Best
Ophthalmologists in America, Chicago Super Docs, Best Doctors, and Top

xxiv
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Doctors. She has authored over 350 articles, 30 book chapters and edited several
books including Age-Related Macular Degeneration (now in the Third Edition). She
has delivered several named lectures, including the inaugural UIC Distinguished
Sweeney Lecture. She has authored or co-authored over 300 articles, 25 book
chapters and edited several books.

Savita Madhusudhan
St Paul’s Eye Unit, Royal Liverpool University Hospital,
Liverpool, UK

Savita Madhusudhan is a consultant ophthalmologist at Royal

Liverpool University Hospital, with special interests in medical
retina, inherited retinal diseases and retinal imaging, and an
honorary senior lecturer at the University of Liverpool. Over
the last 10 years, she has been actively involved in clinical research as co-
investigator and principal investigator for studies in age-related macular degen-
eration and ophthalmic genetics. She is the lead for Liverpool Ophthalmic
Reading Centre, developing and providing retinal image grading and analysis
for various projects both independently and through NetwORC UK. This has
included multimodal grading for hydroxychloroquine retinotoxicity, malarial
retinopathy, studies in age-related macular degeneration and diabetic retinop-
athy, and collaborating on development of automated grading and machine
learning algorithms through the University of Liverpool. She is an associate
editor for the retina section of BMJ Open Ophthalmology.

Olwen C Murphy
Johns Hopkins University School of Medicine, Baltimore, MD,
USA

Dr Olwen C Murphy is a neurology fellow in Johns Hopkins

University School of Medicine. She received her medical degree
from University College Dublin, Ireland. After undertaking resi-
dency training in general internal medicine at Trinity College
Dublin University Hospitals, she completed neurology residency training at Cork
University Hospital, Mercy University Hospital (Cork) and Mater Misericordiae
University Hospital (Dublin), Ireland. Since 2017 she has been a fellow in neurology
at Johns Hopkins University, undertaking specialized training in neuroimmunology
and neurological infections (2017–2020), followed by ocular motor and vestibular
neurology (2020–2021). Her research interests include non-invasive imaging bio-
markers in multiple sclerosis, neurosarcoidosis, and other neuroimmune disorders.

xxv
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Ogugua Okonkwo
Eye Foundation Retina Institute and Eye Foundation Hospital,
Lagos, Nigeria

Dr Ogugua Okonkwo is an ophthalmic and vitreoretinal surgeon. He

has several years of experience in the care of complicated vitreor-
etinal diseases. He has trained several ophthalmologists and vitreor-
etinal fellows and is passionate about advancing the care of
vitreoretinal disease in the African region. He has published several
papers on vitreoretinal-related topics and reviews for a long list of journals, several of
which are indexed on PubMed. He is the editor in chief of two ophthalmic journals, on
the editorial board of several others, and actively mentors juniors in research.

Muhammet Derda Özer

Yuzuncu Yil University Faculty of Medicine, Ophthalmology
Department, Van, Turkey

Dr Lecturer Muhammet Derda Özer studied medicine in Istanbul

University, Istanbul Faculty of Medicine. After completion of his
M.D. degree, he enrolled into the ophthalmology residency pro-
gram at the same Institution. He fulfiled the government duty at
Hakkari State Hospital as a surgeon. He has acquired the postnominal acronym
FICO in 2020 and awarded with a three month fellowship by International Council
of Ophthalmology. He has published a dozen of scientific papers in well-known peer-
reviewed journals. His specific area of interest is ocular immunology and inflamma-
tion, medical and surgical retina. He is currently working as an academic staff
member in Van Yuzuncu Yil University, Faculty of Medicine.

Araniko Pandey
Civil Service Hospital of Nepal, Kathmandu, Nepal

Dr Araniko Pandey completed his MBBS from Nepal Medical

College—Kathmandu University and residency in ophthalmology
from Tribhuvan University. He was an attending ophthalmologist at
Lumbini Eye Institute and Medical Director at Palpa Lions Lacoul
Eye Hospital, a secondary level rural eye care system. He has
published more than a dozen scientific papers and has presented
his research papers at various national and international scientific meetings including,
ARVO and RANZCO.
Dr Pandey is an ECFMG certified ophthalmologist with an experience of
International Research Observership at Bascom Palmer Eye Institute, McKnight
Vision Research Center in the ‘Quantitative Imaging and Eye–Brain Research
Laboratory’, Miami, USA. He is particularly interested in studies of cost effective,
noninvasive, user friendly ophthalmological tools, to identify potential clinical
biomarkers.

xxvi
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Eli Pradhan
Tilganga Institute of Ophthalmology, Kathmandu, Nepal

Dr Eli Pradhan is an ophthalmologist by profession currently

working in Tilganaga Institute of Ophthalmology. Dr Pradhan is
a retina consultant as she is recognized for her skills, knowledge and
empathetic response in this field. Her concern in this field lies not
only to diagnose the patient but she likes to be involved with the
psychological aspects with the treatment. Adding value to the medicine sector,
where doctors are often criticized for too much objectivity and less concern with
patients, she holds the belief that better communication and proper counseling are
very important aspects of any treatment.
She is also the editor in chief of the Nepalese Journal of Ophthalmology and
president of the Nepal Vitreoretinal Society. Compassionate and empathetic,
Dr Pradhan is surely a shining gem in the field of ophthalmology in Nepal.
Despite being recognized as a doctor, her interests include writing and editing in
various journals, newspapers and magazines including Healthy Life. She utilizes her
spare time in making portraits and landscapes, sketching, and is involved in
humanitarian work through organizations like Rotary.

Gunjan Prasai
Kathmandu Medical College, Kathmandu, Nepal

Dr Gunjan completed her MBBS from KIST Medical college,

Tribhuvan University. She is currently a third-year resident in
ophthalmology at Kathmandu medical college and teaching
hospital, Kathmandu Nepal. She has a keen interest in retina
and ocular imaging. She also serves as an editor in Nepalese
Journal of Ophthalmology.

Francesco Prascina
Department of Ophthalmology, University Vita-Salute, IRCCS
Ospedale San Raffaele, Milan, Italy

Francesco Prascina, MD, received his medical degree at the

University of Bari (Italy), and his specialization in ophthalmology
at the Department of Ophthalmology from Bari (Italy). He is
consultant in the Medical Retina and Imaging Unit of the
Department of Ophthalmology, University Vita-Salute, Scientific Institute San
Raffaele (Milan, Italy). His main research interests include the ophthalmic imaging
and management of retinal diseases.

xxvii
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Giuseppe Querques
Department of Ophthalmology, University Vita-Salute, IRCCS
Ospedale San Raffaele, Milan, Italy

Giuseppe Querques, MD, PhD, is Associate Professor at

University Vita-Salute, Scientific Institute San Raffaele, Milan,
Italy. His main topics are medical retina and ophthalmic surgery
(retina and cataract). Professor Querques has contributed to more
than 300 peer-reviewed articles published mainly in the areas of medical retina (age-
related macular degeneration, retinal vascular diseases, hereditary retinal diseases,
ophthalmic genetics). His current main area of both clinical and laboratory research
is focusing on the diagnosis (imaging) and treatment of age-related macular
degeneration, retinal vascular diseases, and hereditary retinal diseases. Doctor
Querques is a member of the Italian Society of Ophthalmology (SOI), the French
Society of Ophthalmology (SFO), the European Association of Retinal Specialists
(EURETINA), the American Academy of Ophthalmology (AAO), the Association
for Research in Vision and Ophthalmology (ARVO), the Retina Society, the Club
Jules Gonin, and the Macula Society. Doctor Querques serves as Editor in Chief of
Ophthalmology @ Point Of Care, and as Editorial Board Member of European
Journal of Ophthalmology, Acta Diabetologica, Ophthalmologica, Ophthalmology
and Therapy, Retinal Cases and Brief Reports, and American Journal Case Report.

Lea Querques
Department of Ophthalmology, University Vita-Salute, IRCCS
Ospedale San Raffaele, Milan, Italy

Lea Querques, MD, received her medical degree and specialization

in ophthalmology at the Department of Ophthalmology from the
University Vita-Salute San Raffaele, Milan (Italy). She is con-
sultant in the Medical Retina and Imaging Unit of the Department
of Ophthalmology, University Vita-Salute, Scientific Institute San Raffaele (Milan,
Italy). Dr Querques has published more than 100 articles in Pub-Med journals
mainly related to retinal diseases. Her main research interests include the ophthalmic
imaging and management of retinal diseases.

Julie Rodman
Nova Southeastern University, Fort Lauderdale, FL, USA

Dr Julie Rodman is the Chief of the Broward Eye Care Institute

in Fort Lauderdale, FL. Her research interests include Optical
Coherence Tomography/Angiography and Vitreoretinal Disease.
Dr Rodman has authored over thirty publications with an emphasis
on retinal disease. She recently published Optical Coherence

xxviii
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Tomography Atlas: A Case Study Approach, the first reference book on this
topic written by an optometrist. Dr Rodman is a member of the American
Optometric Association, American Academy of Optometry, Florida Optometric
Association, Optometric Retina Society and Association for Research in Vision
and Ophthalmology. Dr Rodman is a member of the Optovue Advisory Board
where she serves as a lecturer and consultant. She holds her Oral Pharmaceuticals
Certification and is Laser/Surgical Certified as well.
Dr Rodman has been the recipient of numerous teaching awards, including the
Golden Apple Award for Excellence in Clinical Precepting, and Preceptor of
the Year. She was recognized as a Primary Care Optometry News ‘Top 300’
Optometrists recognizing excellence in innovation and furthering of the profession.
FYI: In her free time, Dr Rodman loves spending time with her husband and two
daughters … just doing anything with them!!!!

Riccardo Sacconi
Department of Ophthalmology, University Vita-Salute, IRCCS
Ospedale San Raffaele, Milan, Italy

Riccardo Sacconi, MD, FEBO, received his medical degree from

the University Vita-Salute San Raffaele, Milan (Italy) in 2012. He
completed his residency in Ophthalmology at University of
Verona (Italy), and a clinical research fellowship in medical retina
and retinal imaging at the department of Ophthalmology of San Raffaele Hospital,
Milan (Italy). Currently, he works as retina specialist at San Raffaele Hospital and
he is completing his PhD program at the Vita-Salute San Raffaele University. His
main research interests include the ophthalmic imaging and management of retinal
diseases, contributing to more than 120 peer-reviewed articles. He is editorial
assistant of European Journal of Ophthalmology and Ophthalmology @ Point
Of Care.

Shiv Saidha
Johns Hopkins University School of Medicine, Baltimore, MD,
USA

Dr Shiv Saidha specializes in the diagnosis, management, and

treatment of multiple sclerosis (MS), as well as other neuro-
immunological disorders of the central nervous system, including
those afflicting the brain, spinal cord, and optic nerves.
Dr Saidha received his medical degree, as well as his postgraduate Doctor of
Medicine, from the National University of Ireland, Galway (NUIG), Ireland. He
completed residency training in general internal medicine at Galway University
Hospitals, Ireland. Subsequent to this, he completed residency training in neurology

xxix
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

at Galway, Cork, and Beaumont (located in Dublin) University Hospitals, Ireland.

He then completed a further three years of specialized fellowship training in
neuroimmunology and neuroinfectious diseases at Johns Hopkins University
School of Medicine, Baltimore, Maryland. He is currently an Associate Professor
of Neurology at Johns Hopkins University School of Medicine.
Dr Saidha’s research interests to date have predominantly focused on the non-
invasive interrogation of retinal structures using optical coherence tomography
(OCT) and OCT angiography (OCTA) in multiple sclerosis (MS), in order to
identify and investigate novel outcome strategies for assessing and monitoring
neuroprotection and neurorestoration in MS, as well as to further our understanding
of the pathobiology of MS, including how retinal pathology in MS may be related to
more global central nervous system disturbances. He has been at the fore of
discovering and demonstrating that inner and outer retinal layer pathology occur
within the unmyelinated retina in MS (both quantitatively and qualitatively). He is
committed to continuing to delineate and ascertain the mechanisms involved in the
development of these findings, the clinical relevance of these changes (both locally in
the retina, as well as more globally), and determine if this information may be used
to guide the development of new, and potentially unique/novel effective MS
therapies. His work to date has been primarily structurally focused (both within
the retina using OCT and OCTA, as well as more globally within the central nervous
system through the assessment of brain substructure volumes, diffusion tensor
imaging and magnetization transfer ratio metrics within specific pathways in MS).
He also has expertise in the functional assessment of the anterior visual pathway in
MS through the application of multifocal electroretinography, pupillometry, and
retinal function imaging techniques, amongst others, as well as of global and
regional brain metabolism. He has published first author papers in Lancet, Lancet
Neurology, Annals of Neurology, Brain, Neurology, JAMA neurology, and Multiple
Sclerosis Journal amongst others.

Muhammad Aamir Shahzad

Aziz Fatimah Medical and Dental College, Faisalabad, Punjab,
Pakistan

Muhammad Aamir Shahzad, MBBS, FRCS, is an Assistant

Professor of Ophthalmology and a Consultant Ophthalmic sur-
geon at Aziz Fatimah Medical and Dental College/Aziz Fatimah
Hospital. He is a dedicated retinal physician and researcher. His
main areas of expertise are epidemiology and management of diabetic retinopathy
and retinal imaging.

xxx
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Kinsuk Singh
Vasan Eye Centre, New Delhi, India

Dr Kinsuk completed her MBBS from Institute of Medicine,

Tribhuvan University, Kathmandu, Nepal. She went on to com-
plete her MD in Ophthalmology from B.P. Koirala Lions Centre
for Ophthalmic Studies, Institute of Medicine, Tribhuvan
University, Kathmandu, Nepal in the year 2016. She completed
her training in anterior segment surgery from Biratnagar Eye Hospital, Biratnagar,
Nepal in the year 2018. She is presently a Member of Royal College of Surgeons of
Edinburgh (MRCSEd). She is currently working at Vasan Eye Care, India.

Taeyoon Son
Department of Biomedical Engineering, University of Illinois at
Chicago, Chicago, IL, USA

Taeyoon Son, PhD is a research assistant professor in the

Biomedical Optics and Functional Imaging lab at the University
of Illinois at Chicago. His primary interest is in the development of
human and animal OCT/OCTA systems. He is currently develop-
ing the technology for oximetry based artery-vein classification in
humans. Dr Son is also leading the project for a portable wide-field fundus camera
for examination of retinopathy of prematurity (ROP). With a strong background in
programming languages such as MATLAB, C, and LabVIEW he has developed
graphical user interfaces for clinical implementation of different algorithms. He has
a PhD in Biomedical Engineering from Yonsei University, Korea.

Meysam Tavakoli
Radiation Oncology Department, University of Texas Southwestern
Medical Center, Dallas, TX, USA

Meysam Tavakoli received the BS degree in physics, Tehran

University, MS in Medical Physics from Mashhad Medical
School, and PhD in physics from Purdue University. Since 2006
he has worked on medical image processing, especially retinal
images. He applied his knowledge toward machine and deep learning for image
processing. His research interests include medical physics, medical image processing,
computer vision, pattern recognition, and statistical models.

xxxi
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Xiyin Wu
College of Big data and Information Engineering, Guizhou
University, Guiyang, China

Xiyin Wu is a lecturer at College of Big data and Information

Engineering, Guizhou University, China. She obtained her PhD
degree in pattern recognition and intelligence systems from
Nanjing University of Science and Technology in 2020. She
obtained her Bachelor’s degree in information security (2012) and the Master’s
degree in computer technology (2015) from Jiangnan University. Her research
interests include image processing, computer vision, pattern recognition and medical
image analysis.

Xincheng Yao
Department of Biomedical Engineering and Department of
Ophthalmology and Visual Sciences, University of Illinois at
Chicago (UIC), Chicago, IL, USA

Xincheng Yao, PhD is a Richard and Loan Hill Professor of

Biomedical Engineering and Ophthalmology and Visual Sciences,
University of Illinois at Chicago (UIC). Dr Yao received his PhD in
optics from the Institute of Physics, Chinese Academy of Sciences in
2001. This was followed by his postdoctoral research in the Biophysical Group at Los
Alamos National Laboratory (LANL) from 2001 to 2004. He held a LANL Technical
Staff appointment from 2004 to 2006 and served at CFD Research Corporation as a
Senior Research Scientist from 2006 to 2007. He was a faculty of University of Alabama
at Birmingham (UAB) from 2007 to 2014 and joined UIC faculty in 2014. Dr Yao’s
research interests include biomedical optics instrumentation, ultra-wide field fundus
photography, functional optical coherence tomography (OCT), OCT angiography,
super-resolution ophthalmoscopy, and machine learning based image classification. He
was elected as a fellow of the International Society for Optics and Photonics (SPIE) and
the American Institute for Medical and Biological Engineering (AIMBE) in 2019.

Yalin Zheng
Eye and Vision Science Department, University of Liverpool,
Liverpool, UK

Yalin Zheng was awarded a PhD in Electronics and Computer

Science from the University of Southampton, Southampton, UK,
in 2003. He is currently Reader in Ophthalmic Imaging at the
Department of Eye and Vision Science, University of Liverpool,
Liverpool, UK. His research interests include image processing, computer vision,

xxxii
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

artificial intelligence and medical image analysis. He has published over 150
conference and journal papers.
Ilaria Zucchiatti
Department of Ophthalmology, University Vita-Salute, IRCCS
Ospedale San Raffaele, Milan, Italy

Ilaria Zucchiatti, MD, received her medical degree at the

University of Udine (Italy), and her specialization in ophthalmol-
ogy at the Department of Ophthalmology, University Vita-Salute
San Raffaele, Milan (Italy). She is consultant in the Medical
Retina and Imaging Unit of the Department of Ophthalmology, University Vita-
Salute, Scientific Institute San Raffaele (Milan, Italy). Her main research interests
include the ophthalmic imaging and management of retinal diseases.

xxxiii
 IOP Publishing

Photo Acoustic and Optical Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
Ayman El-Baz and Jasjit S Suri

Chapter 1
Clinical application of optical coherence
tomography angiography in retinal diseases
Hashim Ali Khan, Muhammad Aamir Shahzad, Muhammad Amer Awan and
Smaha Jahangir

Optical coherence tomography angiography (OCTA) is a functional extension of

optical coherence tomography (OCT) that has made in vivo vascular imaging
possible using motion contrast. Being a recently introduced modality it is rapidly
evolving and so is our understanding of OCTA. In this chapter we review the main
principles behind acquiring ocular blood flow information using OCTA, commerci-
alization of OCTA, and compare it with existing ocular vascular imaging techni-
ques. We discuss clinical interpretation of retinal OCTA, related artifacts and
limitations of the system. Further, OCTA features of important retinal and
choroidal diseases are also reviewed.

1.1 Introduction
In everyday clinical practice, we encounter a significant number of patients where we
need to investigate the retinochoroidal vasculature and hemodyanamics of different
lesions, diagnose the condition and guide treatment plans. The retina is a unique
ocular tissue where the vasculature can be directly visualized non-invasively in vivo.
Most conditions and lesions can be examined clinically in office settings and no
further testing is required. But, oftentimes we may need to study the vascular
properties and hemodynamics of the retinal tissue to differentiate between similar
looking conditions or to tailor our treatment plan. In addition, clinical examination
has its own limitations and many diseases may exist and progress before they
become clinically evident.
Fundus fluorescein angiography (FA) and indocyanine green angiography
(ICGA) are dye based ocular vascular imaging techniques that have been used for
a long time. Retinal vasculature is better visible on FA while ICGA displays the

doi:10.1088/978-0-7503-2060-3ch1 1-1 ª IOP Publishing Ltd 2021

 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

choroidal vasculature more clearly. Both techniques have been the mainstay of
ophthalmic vascular imaging for a long time. However, both are invasive and
associated with significant adverse events.
Optical coherence tomography angiography (OCTA) is a newer and emerging,
non-invasive vascular imaging technique, based on motion contrast extracted from
high speed optical coherence tomography (OCT) images and produces high
resolution, depth resolved images of retinochoroidal vasculature.

1.2 Comparison of OCTA with FA and ICGA

FA and ICGA have been in practice for decades. Both are dye based tests and
require intravenous administration of indocyanine green and fluorescein sodium.
Generally, both agents are considered safe; but they are associated with a significant
number of adverse events and rarely severe reactions and even death. The risk of
adverse effects increases with each repetition, limiting frequent administration of FA
& ICGA. Conversely, OCTA uses motion contrast to extract vascular information
of the tissue under study and is completely non-invasive. It does not cause any
adverse events and can be repeated as many times as required.
During ICGA and FA, we need to study the different phases of dye transit
through the ocular tissue to get useful information. Therefore, these tests last longer
and need significant time for each patient to be studied. OCTA, on the other hand, is
time efficient and gathers vascular images in less than a minute. Both FA and ICGA
are dynamic tests and each phase of dye transit corresponds to a particular vascular
bed at different planes of choroid and retina. Conversely, OCTA is a static test
where each vascular network can be displayed at any desired depth and thickness of
retinochoroidal tissue from volumetric data. In addition, structural information
from corresponding structural OCT scans is also available for any region of interest
on OCT angiograms.
FA and ICGA provide a wide field of view (up to 200 degrees) while OCTA has a
limited field of view typically limited to the posterior pole. However, montaging
multiple OCTA images can produce up to 100-degree field of view (table 1.1). But
with increased field of view the fine details and resolution of image is compromised.
However, multiple OCT angiograms from multiple smaller areas of the retina can be
acquired and combined into one image for extended field of view and with high
detail.
Artifacts are more common in OCTA and are minimal on ICGA or FA. These
may significantly compromise the image quality and details of retinal vasculature.
The common OCTA artifacts include projection artifacts, shadowing and motion
artifacts.

1.3 Principles and methods of OCTA

OCTA is a functional extension of OCT. The conventional OCT devices acquire
structural information of tissue based on partial coherence interferometry. Each
OCT B-Scan consists of several A scans and multiple B Scans are used to reconstruct
en face and three-dimensional structure of the retinal tissue. Each OCT B scan has

1-2
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 1.1. Comparison between FA, ICGA and OCTA.

Parameter FA ICGA OCTA

Invasive Yes Yes No

Label Sodium fluorescein Indocyanine green Label free
Time Needed 10–20 min Up to 35 min Up to 5 min
Ideal for Retinal vasculature Choroidal vasculature Both
Adverse events Relatively common Less common None
Field of view Up to 200 degrees Up to 200 degrees Up to 100 degrees
Artifacts Minimal Minimal Common
Structural information Minimal Minimal Great
Objective features No No Multiple

FA = fluorescein angiography: ICGA = indocyanine green angiography; OCTA = optical coherence

tomography angiography

phase and intensity/ amplitude information of the sample. Based on different

approaches, OCTA uses motion contrast to extract the flow characteristics and
vascular patterns of the tissue of interest. These strategies detect variations in phase,
amplitude or both, caused by moving blood cells against a static background
(retina), from backscattered OCT signals. The methods can be divided into phase-
based methods, amplitude based and complex methods [1–3].

1.3.1 Phase based strategies

The moving blood cell in vasculature causes a phase shift in backscattered OCT
signals. The shift in phase can be used to compute flow measurements. Initial
approaches of phase based OCTA extracted flow information using the phase shifts
between successive A-scans. The initial phase-based approach was employed by
Leitgeb et al [4] in their Doppler OCT (DOCT) using phase information from two
consecutive A-scan lines [2]. Without dedicated scanning protocols and hardware
modifications in DOCT, only axial flow can be measured. Also, the DOCT systems
are subject to smaller motions caused by eye movements. Different attempts have
been made to compensate it. These included histograms based bulk motion
correction between consecutive A-scans and cross correlation of A-lines between
B scan frames. With the continuous advancement in hardware and speed of OCT
systems, the phase information has shifted from comparison of consecutive A scans
to sequential B scans, making detection of slower flow possible. Phase variance
technique by Fingler et al [5–7] using phase analysis between sequential, repeated B
scans at the same location demonstrated vascular imaging comparable to FA. Phase
gradient analysis (PGA) and split-spectrum phase gradient angiography (SSPGA)
are recent phase based methods that map the vasculature without the need for
correction of bulk axial movement. These methods provide a better signal-to-noise
ratio (SNR) and contrast [1–3, 8].

1-3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

1.3.2 Intensity/amplitude based strategies

These approaches use intensity and speckle information from OCT signals. contrary
to phase-based methods, amplitude-based methods are less susceptible to phase noise
[1–3]. During an OCT scan, a random speckle noise is generated as result of OCT
backscattering. However, the speckle patterns generated by moving backscatterers
(moving erythrocytes) is different from static scatterers (the static retinal tissue) [1–3].
Initial approaches included algorithms based on speckle variance. Further improve-
ments were introduced in speckle variance technique by adjusting the number of
repetitions of B scans at a given location [9–11]. Alternative approaches based on
intensity include differential logarithmic intensity variance, intensity difference
between successive B Scans and comparison of vessel reflectivity features from non-
vascular areas [12–14]. Correlation mapping method generates two-dimensional and
three-dimensional flow maps by computing the correlation between successive B scans
[15, 16]. Split-spectrum amplitude-decorrelation (SSADA), described by Jia et al [17]
is an algorithm based on the time-varying speckle. Contrary to correlation mapping
strategies, it computes a decorrelation signal between sequential B scans. The OCT
spectrum is split into narrow bands and each is processed separately generating
multiple flow signals from the same B-scan. The algorithm improves SNR. Due to
reduced axial resolution it is less susceptible to axial bulk motion, making it very
useful for everyday clinical ocular blood flow imaging [2, 17]. OCTA ratio analysis
(OCTARA) is another intensity based proprietary algorithm of Topcon (Topcon
Corporation, Tokyo, Japan) paired with Topcon’s Swept Source (SS) OCT. It is not
based on amplitude decorrelation and the spectrum is not split, thus preserving axial
resolution and optimized for visualization of both retina and choroid [18].

1.3.3 Complex signal-based OCTA

Phase based strategies are less sensitive to transverse flow while intensity-based
approaches have poor performance in detecting slow flow. Complex methods
combine information both from phase and amplitude thus providing better results
[1–3]. Optical microangiography (OMAG), eigen-decomposition of OCT signals
(based on the orthogonality) and the most recent split-spectrum amplitude and
phase gradient analysis (SSAPGA) utilize both phase and amplitude information to
generate vascular flow maps in high resolution. SSAPGA performance is superior to
PGA, SSPGA, and SSADA in terms of contrast and SNR [8].

1.4 Commercial devices

Currently the following OCTA systems are commercially available for ophthalmic
use. Optovue AngioVue™ OCTA System, Topcon Triton SS OCT Angio™, ZEISS
AngioPlex™; Heidelberg Spectralis® OCTA, and Nidek AngioScan.

1.4.1 RTVue XR Avanti

AngioVue is OCTA software deployed on RTVue XR Avanti, a spectral domain (SD)
OCT device. The software is based on SSADA algorithm and uses motion correction

1-4
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

technology (MCT) for correcting motion artifacts and 3D projection artifact (PAR)
removal algorithm to correct projection artifacts. The instrument has an A scan rate
of 70000 per second and utilizes a light source centered on 840 nm with a bandwidth
of 50 nm. The en face OCT angiograms may cover up to an area of 12 × 9 mm.
The angiovue provide many qualitative metrics such as area and perimeter of foveal
avascular zone (FAZ), the flow area and vessel density mapping [19].

1.4.2 Topcon Triton SS OCTAngio™

Topcon’s (Topcon Inc., Tokyo, Japan) OCTA is built into their Topcon Triton SS
OCT based on their proprietary intensity-based algorithm called OCTARA. It has a
speed of 100000 A scans per second and uses a light source of 1050 nm which
provides greater tissue penetration and visualization of choroid without losing
details of vitreous. Scan size is up to 6.0 × 6.0 mm [18].

1.4.3 ZEISS AngioPlex™

Zeiss angioplex (Carl Zeiss Meditec AG, Germany) is an OCT angiography system
based on OMAG algorithm deployed on the CIRRUS™ HD-OCT Model 5000 SD
OCT and PLEX Elite 9000 SSOCT. Cirrus HD OCT 5000 has a light source
centered on 840 nm and has a speed of 68000 A scans per second while PLEX Elite
9000 has a tunable laser centered on 1040 and 1060 nm with a speed of 100000 A
scans per second. Angioplex provides a field of view of up to 8 × 8 mm on Cirrus
5000 and 12 × 12 mm on Plex Elite 9000. The field of view can be further extended
by montaging multiple angiograms [20, 21].

1.4.4 Heidelberg Spectralis® OCTA

This is based on the amplitude decorrelation OCTA algorithm provided on the
Heidelberg’s (Heidelberg Engineering GmbH, Germany) Spectralis SD OCT plat-
form. The device uses a light source of 870 nm and bandwidth of 50 nm with
speed of up to 85000 A scans/second. The field of view for en face angiogram is up to
8.8 × 4.4 mm [22].

1.4.5 Nidek AngioScan

Angioscan is available on the Nidek RS 3000 Advance (Nidek, Aichi, Japan) SD
OCT, which utilizes a light source of 880 nm. The device speed is up to 53000 A
scans per second and it utilizes a modified OMAG algorithm. The field of view is up
to 9 × 9 mm and can be further extended to 12 × 9 by montaging multiple scans. It
can display vascular details of retina and choroid as well as providing vascular
density maps and FAZ area/shape analysis.

1.5 OCTA of normal retina

1.5.1 Retinal and choroidal vasculature
The ocular blood supply is provided by the ophthalmic artery which branches
into the central retinal artery (CRA), long and short posterior ciliary arteries and

1-5
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

anterior ciliary arteries. The choroid is supplied by ciliary arteries which form
arterioles of outer choroid and continuous single layer of wide lumen fenestrated
capillaries of choriocapillaris (CC). The choroid has three vascular networks; the
outer Haller’s layer of larger vessels, inner Sattler’s layer of smaller vessels and CC
which contains fine network of capillaries [23].
The major vascular networks in human retina are superficial vascular plexuses
(SVP) of major retinal arteries, veins and their corresponding branches and capillaries
and is primarily located in the ganglion cell layer (GCL). The two deeper networks,
intermediate capillary plexus (ICP) and deep capillary plexus (DCP) receive their
supply through vertical anastomoses from the SVP. The ICP is a three-dimensional
capillary network at the level of the inner plexiform (IPL) and superficial portion of the
inner nuclear layer (INL) while the DCP is located along a single plane in a deeper
portion of the INL and outer plexiform layer (OPL) [1–3, 24–27].

1.5.2 Default OCTA segmentations

By default, most OCTA devices display SVP, DCP, avascular outer retina (AOR)
and CC as en face angiograms along with corresponding cross-sectional B Scans.
ICP is also displayed by some platforms by default. Nonetheless, segmentations may
be manually adjusted or selected from corresponding B scans to display vascular
networks at desired depth planes [1–3].
The superficial vascular network appears as a homogenous mesh of white lines
(representing vessels) against a dark background (avascular area) distributed across
the entire retina, in a regular centripetal pattern, converged on the FAZ.
The Intermediate capillary plexus has a spider web-like channel and has a well-
defined and smaller FAZ.
The deep capillary plexus appears as a web of fine, densely packed capillaries,
centripetally converged around the perifovea.
AOR segmentation extends from the outer nuclear layer (ONL) to Bruch’s
membrane (BM) and is presented as a dark slab on OCTA.
The CC appears as a homogenous, texture of fine capillaries evenly distributed
across the scan and located between the BM and the choroid. Figure 1.1 shows the
OCTA segmentations and their respective depth locations on structural OCT.

1.5.3 Artifacts
Unlike FA and ICGA, OCTA is typically prone to a wide range of artifacts. These
may be technical artifacts, operator/patient induced artifacts inherent to OCTA
algorithms. OCTA angiograms are acquired using OCT platforms and may suffer
from all artifacts inherent to structural OCT such as movement artifacts, blink
artifacts, segmentation artifacts, etc. Artifacts specific to OCTA are projection
artifacts and shadow artifacts.
Holmen et al [28] graded 406 OCTA images from 234 eyes and it was found that
at least one artifact was present in 97% of the scans. Severe artifacts, associated with
reliability of quantitative measures were present in up to 53%. Shadowing, defocus
and movement were most prevalent in their study.

1-6
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 1.1. Bottom panel shows major layers on structural OCT on the left side and OCTA segmentations on
the right. Top panel shows the color-coded en face angiograms corresponding to SVP, ICP, DCP and CC en
face angiograms. SVP, ICP and DCP copied and edited from [26] under creative common license.

It is necessary to understand the artifacts, their impact on qualitative and

quantitative results on OCT angiograms, before interpreting OCTA results.

1.5.3.1 Motion artifacts

As OCTA generates flow maps using motion contrast strategies, movements of the eye
or within the eye in axial or transverse directions give rise to significant motion artifacts.
The movements in axial direction are usually induced by respiration and cardiac
rhythm. The false flow induced by axial motion is difficult to discern from actual low
flow. However, some algorithms like SSADA are less susceptible to axial bulk motion.
The transverse ocular movements, as a result of saccades and ocular drifts, are
still a problem. The movements appear as white lines on en face angiograms. Many
manufactures use eye-tracking and algorithms to correct for transverse motion. But
often these corrections, gives rise to another set of artifacts such as vessel doubling,
stretching and crisscrossing.
Ocular blinks momentarily block the OCT signals and can be identified as
horizontal black lines on en face angiograms. Figure 1.2 shows blink and transverse
motion artifacts.

1.5.3.2 Shadow / masking artifacts

OCT signal attenuation from generalized media opacities (corneal disorders,
cataracts and vitreous opacities) and poor beam alignment may result in low
contrast and false low flow or flow voids on angiograms. Most devices display an

1-7
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 1.2. Black horizontal lines (arrow) representing blink artifacts and vertical faint white lines (arrow
head) corresponding to transverse motion artifacts.

Figure 1.3. Shadowing from vitreous visible on en face OCT and OCTA (all slabs) and from intraretinal
exudates to most posterior angiograms.

SNR which may be taken into account while interpreting the scans. A high SNR is
always desired.
Vitreous floaters, tufts of vitreous hemorrhage, inflammatory cells, etc may
interfere with the OCT signals resulting in localized signal attenuation and
corresponding pseudo low flow on OCT angiograms. Side by side evaluation of
OCT shadowgram, en face structural and cross-sectional images are helpful in
differentiating true low flow from false low flow [2]. Figure 1.3 shows shadow
artifacts.

1-8
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

1.5.3.3 Segmentation artifacts

Several approaches are used to identify retinal layer boundaries in OCT to generate
en face angiograms of desired layers. These include layer reflectivity, texture, use of
deep learning, etc. Layer continuity is frequently employed implied in segmentation
strategies. Usually, these algorithms demonstrate excellent performance towards
identifying the layer boundaries in normal retinae. However, the algorithms
frequently fail segmenting the boundaries accurately in diseased eyes. Eyes with
macular edema, choroidal neovascularization, high myopia, preretinal membrane
and retinal traction pose a special challenge to segmentation algorithms [29].
On en face angiograms, the given slab may have information from other layers.
Concomitant cross-sectional image analysis and manual correction of segmentation
can help in such cases; but, it is a tiring and time-consuming job to correct
segmentations across multiple B Scans.

1.5.3.4 Projection artifacts

These artifacts are by far the most common and inherent to all OCTA scans. When
an OCTA beam passes through blood vessels, it may be reflected back, scattered,
absorbed or refracted. The OCTA beam passing through these blood vessels and
reaching deeper layers suffers time varying fluctuations. The fluctuating beam,
subsequently, highlights the posterior structure. Some of these structures including
retinal pigment epithelium (RPE), plexiform layers and choriocapillaris act as
reflective surfaces. The reflection of this, time varying fluctuating signal across
B-scans is, falsely, computed as flow signal coming from the reflecting surfaces and
displayed as duplication of superficial flow maps on deeper slabs. Small reflective
structures, such as drusen, exudates and pigment can also give rise to localized
projection artifacts.
Several methods have been described to correct projection artifacts. Commonly
employed is subtraction of superficial flow maps from deeper slabs; but it gives rise
to another troublesome artifact. The pattern of bright vessels is replaced by dark
lines. At present, no method can effectively correct these artifacts.
Clinically, uncorrected projection artifacts appear as duplication of superficial
flow maps on deeper slabs and duplication of superficial vascular patterns as dark
lines on posterior structures suggest a corrected projection artifact. Figure 1.4 shows
projection artifacts.

1.6 OCTA in retinal vascular diseases

OCTA, being a non-invasive vascular imaging method, is increasingly finding its
role in diagnosis in management of retinal vascular diseases. It can also be used as a
standalone vascular imaging technique in most cases and as adjunct to FA in some
situations. Management of most of the vascular diseases of the retina needs
structural OCT information for establishing the diagnosis, severity grading and
decisions to treat. OCTA being available on the same platform can be acquired
quickly and studied alongside structural OCT scans.

1-9
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 1.4. Projection of SVP vessels (indicated by white arrow) on DCP, outer retina and CC.

1.6.1 Diabetic retinopathy

OCTA has comparable results to FA in demonstrating the classical clinical findings
of diabetic retinopathy (DR) [2, 30]. There exists a great agreement between FA and
OCTA in measurement of the FAZ diameter, a potential new disease biomarker, as
well as in detecting microaneurysms, vascular beading, capillary non-perfusion
(CNP) and vascular proliferation [30, 31]. Nonetheless, OCTA is non-sensitive to
vascular leakage, hence it may display the actual extent of retinal neovascularization
which is obscured by leakage on FA [2].
OCTA provides multiple various qualitative metrics for assessing and monitoring
DR including vessel density, intercapillary distance, measurement of CNP area at
different vascular networks of the retina. Rosen et al [32] suggested higher perfused
capillary density (PCD) as a sign of preclinical diabetic retinopathy and that it
warrants the use of OCTA in early detection of DR. Incessant capillary loss in DR
causes a low PCD in eyes with clinical diabetic retinopathy. As FA displays vascular
information as a composite map of all vascular beds, we cannot visualize each
individual vascular tree. OCTA allows us to view individual as well composite flow
maps and provides better contrast due to absence of background choroidal flush.
Figure 1.5 shows OCTA findings in diabetic retinopathy.

1.6.2 Retinal vein occlusion (RVO)

The severity of visual symptoms and prognosis of both branch and central RVO
depends on the severity of macular edema and extent of vascular insufficiency. OCT
is the most common test used for guiding treatment in eyes with venous occlusions.
OCTA can delineate the area of vascular insufficiency, CNP, FAZ changes and
other potential complications of RVO can be monitored [33–36]. Nonetheless, due
to marked macular edema in venous occlusions, correct segmentation of retinal slabs
often becomes a challenge. However, the composite full thickness OCTA maps can
be used to obtain useful vascular flow information [37].

1-10
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 1.5. Multiple microaneurysms, irregular and widened foveal avascular zone and capillary dropout in
SVP and DCP. Projection of SVP vessels and localized shadowing on CC can be seen.

Studies of venous occlusions, using OCTA, have demonstrated dilated venous

segment within areas of retinal edema. Furthermore, it was found that vascular
insult was more profound in DCP compared to SVP [38–41]. Figure 1.6 shows
dilated veins, flow deficits and macular edema in the case of central retinal vein
occlusion.
OCTA has proven to be comparable to FA in documenting vascular changes in
RVO, especially when using 3 × 3 mm and HD angio scans. However, with
increasing field of view the performance is reduced. Peripheral vascular non-
perfusion is a common feature in eyes with RVO and is an important prognostic
factor. The smaller field of view on OCTA limits its utility in documenting
peripheral ischemia [2, 3].

1.6.3 Retinal artery occlusion

Retinal artery occlusion (RAO) manifests as occlusion of one of the branches,
hemiretinal arteries or central retinal artery secondary to thrombosis, arterial
embolism or increased intraocular pressure. The severity of occlusion ranges from
partial reduction to complete loss of arterial flow.
A limited number of case reports and series about OCTA findings in RAO have
been published. Overall, good agreement exists between FA and OCTA in studying
the vascular status in RAO [2].
de Castro-Abeger et al [42] demonstrated reduced flow in the area distal to the
embolus in branch artery occlusion. In their study, flow reduction was more
profound in corresponding SVP compared to DCP. Other studies have reported

1-11
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 1.6. Dilated veins with large flow voids and macular edema in the case of retinal vein occlusion.

the severe perfusion defects in both superficial and deep vascular networks in acute
cases and severity of low perfusion substantially improving over time in DCP in
chronic cases [43, 44].
Optic disc perfusion may be relatively preserved in acute cases progressing to
diffuse attenuation of radial peripapillary capillaries in chronic phase [2, 45, 46].

1.6.4 Central serous chorioretinopathy

Central serous chorioretinopathy (CSCR) is mostly a self-limiting retinal disease
characterized by serous separation of the neurosensory retina. A relatively small
proportion of CSCR becomes chronic in course and may further complicate by
developing choroidal neovascularization (CNV). Often, chronic cases are difficult to
distinguish from age-related macular degeneration and other causes of CNV.
OCTA in acute stages shows a normal flow profile in both SVP and DCP.
However, CC textural alteration and flow impairments, either due to impaired flow

1-12
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

or shadowing from overlying subretinal fluid, have been described in acute stages
[47, 48].
Further, OCTA may be a great tool in monitoring chronic CSCR for develop-
ment of CNV and differentiating it from similar looking diseases. OCTA showed a
high sensitivity in detecting CNV in eyes with CSCR [2].

1.6.5 Macular telangiectasia

Macular telangiectasia (MacTel) is an acquired vascular disorder of retinal
vasculature in the elderly. The clinical manifestations include loss of foveal trans-
lucency in earliest stages. With the advancement of disease there may be retinal
cavitation, volume loss, vascular alterations, right angled venules, intraretinal
crystals, RPE pigmentations and vascular remodeling. It may further progress to
and is complicated by CNV formation [49].
The earlier clinical findings may be very subtle to detect clinically and may be
missed. FA is the current gold standard in diagnosing MacTel. It highlights the
telangiectatic vasculature in early phases of angiogram followed by diffuse hyper-
fluorescence in later phases [49].
Structural OCT is very useful in evaluating the retinal cavitation, photoreceptor
disruptions, changes in retinal reflectivity and potential development of CNV [49].
Studies have shown great agreement between FA and OCTA in exhibiting the
features of MacTel [2].
OCTA shows dilatation of temporal juxtafoveal capillaries of DCP in early stages
of the disease. The photoreceptor disruption on OCT corresponds to the area of
vascular alterations in DCP [49–56].
Multiple aneurysmal dilatations of DCP vessels extending to outer retina with a
concordant area of ellipsoid zone loss on structural OCT are characteristics of the
intermediate stage [49–56].
In proliferative stage, severe vascular abnormalities are seen in both SVP and
DCP. Vascular anastomosis is displayed as abnormal vessels in outer retinal slabs.
Submacular CNV can be seen in this stage [49–56]. Figure 1.7 shows OCTA
displaying changes in early and advanced macular telangiectasia.
Rarefaction of retinal capillaries, widened intercapillary distances and reduction
in capillary density, both in SVP and DCP have been demonstrated in multiple
studies [49–56]. In a study by Toto et al [49] parafoveal capillary density in MacTel
was reduced in both DCP and SVP. However, the difference between MacTel and
normal eyes was significant only in SVP.

1.6.6 Other retinal diseases

OCTA may be useful in diagnosing and managing a variety of other retinal diseases
including retinal arterial macroaneurysms, radiation retinopathy, Coats’ disease,
foveal hypoplasia, Purtscher’s retinopathy, retinitis pigmentosa, Stargardt disease,
best vitelliform dystrophy, etc. Not only can it aid in diagnosis but it also delivers
additional insights into our understanding of these diseases by providing details of
each individual vascular network in retina and choroid [2].

1-13
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 1.7. Composite en face OCTA and Flow B scan (a) of a case with early macular telangiectasia
demonstrating microaneurysms in temporal juxtafovea. OCTA of and advanced case of MacTel (b) with
vascular remodeling and irregular FAZ on OCTA and volume loss on B-Scan.

1.7 Choroidal diseases

1.7.1 Dry age-related macular degeneration and variants
Age-related macular degeneration (AMD) is the most common retinal degeneration
in the elderly. Drusen under RPE with disruptions/alteration of overlying photo-
receptors are the earliest sign. The disease further progresses to pigmentary
abnormalities, more marked photoreceptor loss and geographic atrophy (GA) of
chorioretinal tissue. CC inadequacy has been implicated as a mechanism behind
development of drusen [57, 58].
OCTA has demonstrated flow impairment in CC under the drusen. However, it is
unclear if it is true flow impairment or the shadowing/signal attenuation caused by
drusen leads to a false low flow. SS OCTA have demonstrated lesser signal loss
compared to SD OCTA. Yet, it is believed that CC flow impairment in part may be
contributed by signal attenuation and imparted by true flow impairment [59–62].
CC flow alterations and capillary rarefaction under and beyond the margins of GA
have also been described using conventional OCTA and the variable interscan time
analysis (VISTA) algorithm, which is particularly helpful in differentiating true low
flow from false low flow. Regardless of any system and algorithms used CC under

1-14
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

drusen and GA demonstrated flow impairment. Reduced vascular density in SVP and
DCP of eyes with early and moderate AMD have also been reported [62–64].
Figure 1.8 shows features of geographic atrophy on OCTA and structural OCT.
Although GA in advanced dry AMD can be well visualized clinically and on
structural OCT, OCTA can be used to delineate the extent of GA and study the

Figure 1.8. severe retinal thinning on OCT and complete absence of choriocapillaris on OCTA in a case of
GA. Larger choroidal vessels are visible on CC slabs due to increased OCT signal penetration.

1-15
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

alteration in CC and deeper choroidal vessels (visible through CC defects).

At this stage research has demonstrated reduced vascular density in both SVP
and DCP [65].
In addition, OCTA offers further benefits by resolving individual vascular tissue
in the retina which is not possible with FA and ICGA. CC, in AMD and other
conditions, can be better resolved and visualized using OCTA than with FA or
ICGA.

1.7.2 Wet AMD and variants

CNV is the most important feature of wet AMD. Typing of CNV, differentiation
between CNV variants and early intervention may yield better results and visual
outcomes. Other features of wet AMD include pigment epithelial detachment
(PED), subretinal hemorrhaging, RPE tears, retinal edema and subsequent chorior-
etinal scarring and irreversible visual loss.
FA and ICGA are often used to diagnose and differentiate between types of CNV
and its variants, while OCT is routinely employed to guide treatment decisions.
Origin and anatomic relationships of CNV are important in differentiating
between the three types of CNV. A less distinct FA lesion located and limited to
sub RPE space is Type I (or occult) CNV, whereas type II (classic) CNV has
proliferated onto and grows in subretinal space. Type III CNV, or retinal
angiomatous proliferation (RAP) originates intraretinally and manifests with intra-
retinal hemorrhaging, exudation and edema. Anastomosis between intraretinal and
subretinal components may be with type III lesions [66].

1.7.2.1 Type I CNV

It is often difficult to detect on OCT and has an indistinct appearance on FA. ICGA
may be required in some cases to diagnose type I CNV. On OCTA, it has a
distinctive appearance as a highly vascular complex of a pruned vascular tree or
vascular loop or a tangled network on CC slabs. A hyperreflective halo surrounding
the neovascular tree is often seen which may be due to shadowing from RPE or more
likely a CC hypoperfusion defect [1–3].
OCTA may have a sensitivity similar to or better than FA, ICGA and OCT in
detecting type I CNV. However, CNV size may appear smaller on OCTA compared
to ICGA, likely because it detects the only flow column in CNV, while in ICGA the
vascular walls as well as lumen are both visualized [2, 67, 68].

1.7.2.2 Type II CNV

A glomerular or medusa patterned hyperflow lesion is seen in outer avascular retina
slabs of OCT angiograms. A thicker feeder trunk from the neovascular complex
communicates with the deeper choroid is usually seen. Similar to type I CNV, a
prelesional hypoflow halo is frequently visible [1–3]. Figure 1.9 shows a very subtle
small type II CNV visible on CC and AOR slabs. The lesion was barely perceptible
on structural OCT on careful inspection.

1-16
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 1.9. A very small choroidal neovascular complex is visible in the outer retina and CC (top row). The
lesion can be barely noticed on careful inspection of OCT B scan (bottom row).

OCTA may be a great tool in baseline assessment of CNV as well as guiding

us throughout the treatment course of CNV, although some clinicians are
more comfortable with FA/ICGA for guiding treatment response and decision.
Nonetheless, unlike FA/ICGA, OCTA only highlights CNV with patent flow
(active part) along with structural information from reflectivity profiles, so we may
use it confidently. However, in certain cases FA/ ICGA are required to better

1-17
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 1.10. A large subretinal hyperreflective lesion on OCT (bottom row) with RPE rip and cystic changes in
retina. The neovascular complex can be seen in outer retinal and CC slabs. Multiple artifacts including
segmentation artifacts and projection artifacts are noticeable.

understand the nature and activity of lesions. Figure 1.10 shows advanced type II
CNV with retinal edema, RPE rip and cystic changes. Segmentation errors are
apparent.

1.7.2.3 Type III CNV

Retinal angiomatous proliferation is an AMD variant which may be difficult to
distinguish from classic CNV in some cases.
Type III lesions appear as a hyperflow signal in outer retinal slabs as a tuft of
small, curvilinear vessels anastomosing DCP through a feeder vessel. A small,
corkscrew like lesion in CC slab is commonly seen.

1.7.2.4 Polypoidal choroidal vasculopathy

Polypoidal choroidal vasculopathy (PCV) manifests as bilateral, multiple, recurrent
serosanguineous and serous RPE detachments secondary to branching vascular
network (BVN) and polypoidal lesions in choroid [66, 69].
Polyps appear as round hyperflow lesions in a nodular, ring or clustered
conformation in CC segmentations. The central lesion is usually surrounded by a
dark halo, while BVN assumes configuration of medusa or seafan of tangled bright
vascular signal in CC.
Compared to ICGA, BVN may be better visualized on OCTA; however, ICGA is
superior in detecting and delineating polyps [1–3].
At present, OCTA may not replace the current gold standard vascular imaging
techniques for CNV diagnosis and management. Nevertheless, it provides us with a

1-18
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

unique opportunity to study the changes in each individual vascular network

independently, furthering our understanding of these vascular lesions. Further, the
availability of structural OCT information, in addition to flow characteristics equips
us with more information compared to FA or ICGA alone.

1.8 OCTA in ocular inflammation

OCTA is increasingly finding its role not only in the diagnosis and management of
posterior uveitides but also in expanding our understanding of the disease processes.
The common use of FA, ICGA and structural OCT in ocular inflammation
includes documentation of macular edema, vascular leakage, vasculitis, retinal
ischemia, choroidal neovascularization and differentiation between similar looking
conditions.
Retinal vasculitis, a common sign in a variety of ocular and systemic diseases, is
characterized by diffuse/segmental inflammation and leakage of retinal arteries and
veins which further progresses to capillary non perfusion, retinal neovascularization
and vitreous hemorrhage. OCTA, although, cannot detect leakage from inflamed
vessels that is well displayed by FA; it highlights the actual changes in vasculature
because the view is not obscured by dye leakage, unlike FA. It further, helps in
delineating the areas of non-perfusion and retinal vascular proliferation. OCTA has
been utilized to study the features of many retinal vasculitis spectra including
Idiopathic Retinal Vasculitis, Aneurysms and Neuroretinitis (IRVAN) syndrome. In
IRVAN eyes it can detect retinal aneurysms with a sensitivity similar to FA.
However, most cases of this disease spectrum are associated with peripheral CNP
which can be missed by OCTA due to smaller field of view. Montaging multiple
images from different fields can be used to generate OCT angiograms with wider
field of view.
In many of the posterior uveatides, hypofluorescence of inflammatory foci in
early phases of FA are typical. The origin of the hypofluorescence and hypocya-
nance on ICGA has been long debated to originate from the CC hypoperfusion or
choroidal masking/ shadowing from the inflammatory foci or cloudy swelling of
retinal pigment epithelium. OCTA also has confirmed the dark spots on CC slabs
corresponding to true low flow or signal attenuation because of inflammatory foci.
The CC perfusion defect is believed to be true flow void because, on co-registered
B-Scans and en face images, the CC perfusion defect sizes were larger than the focus
of inflammation. Further, alterations in CC texture have been described in a number
of eyes after complete resolution of inflammation in primary inflammatory chorio-
capillaropathies and Vogt–Koyanagi–Harada’s disease [70–73].
Paracentral acute middle maculopathy (PAMM), poses a clinical challenge for the
diagnosis. It is characterized by parafoveal hyperreflective band at the level of OPL to
IPL, on structural OCT, in the absence of any FA correlates. OCTA, however, has
revealed perfusion abnormalities in both SVP and DCP. The hypoperfusion, as may
be expected, is more marked in DCP. The visual deficit in PAMM corresponds well
with the severity and location of non-perfusion on OCTA. The persistent flow defects
on OCTA explain the progressive thinning of retina in PAMM [2, 70].

1-19
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

1.9 Conclusion
FA and ICGA have remained the gold standard of vascular imaging for decades.
However, associated adverse events were a major factor limiting their frequent use.
OCTA is an emerging vascular imaging modality based on motion contrast capable
of depth resolved imaging of ocular vasculature using motion contrast, non-
invasively.
It has shown excellent results comparable to FA and ICGA in most of the
situations. Its unique capability of resolving individual vascular networks provides
more information about the origin of a disease process as well as tracing the vascular
beds from where a particular disease process actually begins.
Despite its non-invasive nature, high resolution, excellent reproducibility, OCTA
algorithms are prone to a significant number of different artifacts limiting their
usefulness in everyday clinical practice. A good understanding of these artifacts and
their impact on the displayed angiograms is necessary to derive useful information.
Even though, OCTA has broadened our understanding about ocular disease, its
clinical usefulness and applicability is still evolving; we recommend it as a supple-
ment to existing vascular imaging instead of using it as an exclusive tool. With future
improvements in scan acquisition speeds, improved hardware and newer OCTA
algorithms we expect to obtain more useful information with fewer artifacts in
shorter time.

References
[1] Kashani A H, Chen C L and Gahm J K et al 2017 Optical coherence tomography
angiography: A comprehensive review of current methods and clinical applications Prog.
Retin. Eye Res 60 66–100
[2] Khan H A, Mehmood A and Khan Q A et al 2017 A major review of optical coherence
tomography angiography Expert Rev. Ophthalmol 12 373–85
[3] Spaide R F, Fujimoto J G and Waheed N K et al 2018 Optical coherence tomography
angiography Prog. Retin. Eye Res 64 1–55
[4] Leitgeb R A, Werkmeister R M and Blatter C et al 2014 Doppler optical coherence
tomography Prog. Retin. Eye Res 41 26–43
[5] Fingler J, Readhead C and Schwartz D M et al 2008 Phase-contrast OCT imaging of
transverse flows in the mouse retina and choroid Investig. Ophthalmol. Vis. Sci. 49 5055–59
[6] Fingler J, Schwartz D and Yang C et al 2007 Mobility and transverse flow visualization
using phase variance contrast with spectral domain optical coherence tomography Opt.
Express 15 12636–53
[7] Fingler J, Zawadzki R J and Werner J S et al 2009 Volumetric microvascular imaging of
human retina using optical coherence tomography with a novel motion contrast technique
Opt. Express 17 22190–200
[8] Liu G, Jia Y and Pechauer A D et al 2016 Split-spectrum phase-gradient optical coherence
tomography angiography Biomed. Opt. Express 7 2943–54
[9] Barton J and Stromski S 2005 Flow measurement without phase information in optical
coherence tomography images Opt. Express 13 5234–9
[10] Mariampillai A, Standish B A and Moriyama E H et al 2008 Speckle variance detection of
microvasculature using swept-source optical coherence tomography Opt. Lett. 33 1530–2

1-20
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[11] Mariampillai A, Leung M K and Jarvi M et al 2010 Optimized speckle variance OCT
imaging of microvasculature Opt. Lett. 35 1257–9
[12] Motaghiannezam S M, Koos D and Fraser S E 2012 Differential phase-contrast, swept-
source optical coherence tomography at 1060 nm for in vivo human retinal and choroidal
vasculature visualization J. Biomed. Opt. 17 026011
[13] Blatter C, Klein T and Grajciar B et al 2012 Ultrahigh-speed non-invasive widefield
angiography J. Biomed. Opt. 17 070505
[14] Schmoll T, Singh A S and Blatter C et al 2011 Imaging of the parafoveal capillary network
and its integrity analysis using fractal dimension Biomed. Opt. Express 2 1159–68
[15] Enfield J, Jonathan E and Leahy M 2011 In vivo imaging of the microcirculation of the volar
forearm using correlation mapping optical coherence tomography (cmOCT) Biomed. Opt.
Express 2 1184–93
[16] Jonathan E, Enfield J and Leahy M J 2011 Correlation mapping method for generating
microcirculation morphology from optical coherence tomography (OCT) intensity images
J. Biophotonics 4 583–7
[17] Jia Y, Tan O and Tokayer J et al 2012 Split-spectrum amplitude-decorrelation angiography
with optical coherence tomography Opt. Express 20 4710–25
[18] Stanga P E, Tsamis E and Papayannis A et al 2016 Swept-source optical coherence
tomography angio (Topcon Corp, Japan): technology review Dev. Ophthalmol 56 13–7
[19] Huang D, Jia Y and Gao S S et al 2016 Optical coherence tomography angiography using
the Optovue device Dev. Ophthalmol 56 6–12
[20] ZEISS CIRRUS OCT with AngioPlex – Making the revolutionary, routine. – ZEISS
Medical Technology: ZEISS International [2 May 2020]. Available from https://zeiss.com/
meditec/int/specialties/retina/diagnostics/optical-coherence-tomography/oct-optical-coherence-
tomography/cirrus-oct-with-angioplex.html
[21] Rosenfeld P J, Durbin M K and Roisman L et al 2016 ZEISS Angioplex™ spectral domain
optical coherence tomography angiography: technical aspects Dev. Ophthalmol 56 18–29
[22] Rocholz R, Teussink M M and Dolz-Marco R et al 2018 SPECTRALIS Optical Coherence
Tomography Angiography (OCTA): Principles and Clinical Applications https://heidelber-
gengineering.com/media/e-learning/Totara/Dateien/pdf-tutorials/210111-001_SPECTRALIS
%20OCTA%20-%20Principles%20and%20Clinical%20Applications_EN.pdf
[23] Nickla D L and Wallman J 2010 The multifunctional choroid Prog. Retin. Eye Res. 29
144–68
[24] Tan P E Z, Yu P K and Balaratnasingam C et al 2012 Quantitative confocal imaging of the
retinal microvasculature in the human retina Investig. Ophthalmol. Vis. Sci. 53 5728–36
[25] Tan P E Z, Yu P K and Cringle S J et al 2014 Quantitative assessment of the human retinal
microvasculature with or without vascular comorbidity Investig. Ophthalmol. Vis. Sci. 55
8439–52
[26] Hirano T, Chanwimol K and Weichsel J et al 2018 Distinct retinal capillary plexuses in
normal eyes as observed in optical coherence tomography angiography axial profile analysis
Sci. Rep. 2018/06/20 8 9380
[27] Campbell J P, Zhang M and Hwang T S et al 2017 Detailed vascular anatomy of the human
retina by projection-resolved optical coherence tomography angiography Sci. Rep. 7 42201
[28] Holmen I C, Konda S M and Pak J W et al 2020 Prevalence and severity of artifacts in
optical coherence tomographic angiograms JAMA Ophthalmol 138 119–26

1-21
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[29] Spaide R F and Curcio C A 2017 Evaluation of segmentation of the superficial and deep
vascular layers of the retina by optical coherence tomography angiography instruments in
normal eyes JAMA Ophthalmol 135 259–62
[30] Khadamy J, Abri Aghdam K and Falavarjani K G 2018 An update on optical coherence
tomography angiography in diabetic retinopathy J. Ophthalmic. Vis. Res 13 487–97
[31] Enders C, Baeuerle F and Lang G E et al 2020 Comparison between findings in optical
coherence tomography angiography and in fluorescein angiography in patients with diabetic
retinopathy Ophthalmologica. 243 21–6
[32] Rosen R B, Andrade Romo J S and Krawitz B D et al 2019 Earliest evidence of preclinical
diabetic retinopathy revealed using optical coherence tomography angiography perfused
capillary density Am. J. Ophthalmol. 203 103–15
[33] Kuehlewein L, An L and Durbin M K et al 2015 Imaging areas of retinal nonperfusion in
ischemic branch retinal vein occlusion with swept-source OCT microangiography
Ophthalmic. Surg. Lasers Imaging Retina 46 249–52
[34] Nobre Cardoso J, Keane P A and Sim D A et al 2015 Systematic evaluation of optical
coherence tomography angiography in retinal vein occlusion Am. J. Ophthalmol. 163 93-107.
e6
[35] Rispoli M, Savastano M C and Lumbroso B 2015 Capillary network anomalies in branch
retinal vein occlusion on optical coherence tomography angiography Retina 35 2332–8
[36] Sakimoto S, Gomi F and Sakaguchi H et al 2015 Analysis of retinal nonperfusion using
depth-integrated optical coherence tomography images in eyes with branch retinal vein
occlusion Invest. Ophthalmol. Vis. Sci. 56 640–6
[37] Novais E A and Waheed N K 2016 Optical coherence tomography angiography of retinal
vein occlusion OCT Angiography Retin. Macular Diseases vol 56 (Basel: Karger Publishers)
132–38
[38] Adhi M, Filho M A B and Louzada R N et al 2016 Retinal capillary network and Foveal
avascular zone in eyes with vein occlusion and fellow eyes analyzed with optical coherence
tomography angiographyretinal capillary network and FAZ in RVO with OCTA Investig.
Ophthalmol. Vis. Sci. 57 OCT486–94
[39] Casselholmde Salles M, Kvanta A and Amrén U et al 2016 Optical coherence tomography
angiography in central retinal vein occlusion: correlation between the Foveal avascular zone
and visual acuityOCT angiography in CRVO Investig. Ophthalmol. Vis. Sci. 57 OCT242–6
[40] Coscas F, Glacet-Bernard A and Miere A et al 2016 Optical coherence tomography
angiography in retinal vein occlusion: evaluation of superficial and deep capillary plexa
Am. J. Ophthalmol. 161 160–71
[41] Tsuboi K, Ishida Y and Kamei M 2017 Gap in capillary perfusion on optical coherence
tomography angiography associated with persistent macular edema in branch retinal vein
occlusion Investig. Ophthalmol. Vis. Sci. 58 2038–43
[42] de Castro-Abeger A H, de Carlo T E and Duker J S et al 2015 Optical coherence
tomography angiography compared to fluorescein angiography in branch retinal artery
occlusion Ophthalmic Surg. Lasers Imaging Retina 46 1052–4
[43] Yu S, Pang C E and Gong Y et al 2015 The spectrum of superficial and deep capillary
ischemia in retinal artery occlusion Am. J. Ophthalmol. 159 53–63
[44] Bhanushali D R, Yadav N K and Dabir S et al 2016 Spectral domain optical coherence
tomography angiography features in a patient of central retinal arterial occlusion before and
after paracentesis Retina 36 e36–8

1-22
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[45] Bonini Filho M A, Adhi M and de Carlo T E et al 2015 Optical coherence tomography
angiography in retinal artery occlusion Retina 35 2339–46
[46] Lee A Y, Zhang Q and Baughman D M et al 2016 Evaluation of bilateral central retinal
artery occlusions with optical coherence tomography-based microangiography: a case report
J. Case Rep 10 307
[47] Aggarwal K, Agarwal A and Deokar A et al 2017 Distinguishing features of acute Vogt–
Koyanagi–Harada disease and acute central serous chorioretinopathy on optical coherence
tomography angiography and en face optical coherence tomography imaging J. Ophthalmic
Inflamm. Infect 7 3
[48] Costanzo E, Cohen S Y and Miere A et al 2015 Optical coherence tomography angiography
in central serous chorioretinopathy J. Ophthalmol 2015 134783
[49] Toto L, Di Antonio L and Mastropasqua R et al 2016 Multimodal imaging of macular
telangiectasia Type 2: focus on vascular changes using optical coherence tomography
angiography imaging macular telangiectasia type 2 by OCTA Invest. Ophthalmol. Vis. Sci.
57 OCT268–76
[50] Chidambara L, Gadde S G and Yadav N K et al 2016 Characteristics and quantification of
vascular changes in macular telangiectasia type 2 on optical coherence tomography
angiography Br. J. Ophthalmol. 100 1482–8
[51] Engelbert M and Yannuzzi L A 2012 Idiopathic macular telangiectasia type 2: the
progressive vasculopathy Eur. J. Ophthalmol 23 1–6
[52] Gaudric A, Krivosic V and Tadayoni R 2015 Outer retina capillary invasion and ellipsoid
zone loss in macular telangiectasia type 2 imaged by optical coherence tomography
angiography Retina 35 2300–6
[53] Roisman L and Rosenfeld P 2016 Optical coherence tomography angiography of macular
telangiectasia type 2 OCT Angiography Retin. Macular Diseases vol 56 (Basel: Karger
Publishers) 146–58
[54] Spaide R F, Klancnik J M Jr and Cooney M J 2015 Retinal vascular layers in macular
telangiectasia type 2 imaged by optical coherence tomographic angiography JAMA
Ophthalmol 133 66–73
[55] Spaide R F, Klancnik J M Jr and Cooney M J et al 2015 Volume-rendering optical
coherence tomography angiography of macular telangiectasia type 2 Ophthalmology 122
2261–9
[56] Thorell M R, Zhang Q and Huang Y et al 2014 Swept-source OCT angiography of macular
telangiectasia type 2 Ophthalmic Surg. Lasers Imaging Retina 45 369–80
[57] Mullins R F, Johnson M N and Faidley E A et al 2011 Choriocapillaris vascular dropout
related to density of drusen in human eyes with early age-related macular degeneration
Invest. Ophthalmol. Vis. Sci. 52 1606–12
[58] Lengyel I, Tufail A and Hosaini H A et al 2004 Association of drusen deposition
with choroidal intercapillary pillars in the aging human eye Invest. Ophthalmol. Vis. Sci.
45 2886–92
[59] Chu Z, Gregori G and Rosenfeld P J et al 2019 Quantification of choriocapillaris with
optical coherence tomography angiography: a comparison study Am. J. Ophthalmol. 208
111–23
[60] Chatziralli I, Theodossiadis G and Panagiotidis D et al 2018 Choriocapillaris vascular
density changes in patients with drusen: cross-sectional study based on optical coherence
tomography angiography findings Ophthalmol. Ther 7 101–7

1-23
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[61] Lane M, Moult E M and Novais E A et al 2016 Visualizing the choriocapillaris under
drusen: comparing 1050-nm swept-source versus 840-nm spectral-domain optical coherence
tomography angiography Invest. Ophthalmol. Vis. Sci. 57 OCT585–90
[62] Moreira-Neto C A, Moult E M and Fujimoto J G et al 2018 Choriocapillaris loss in
advanced age-related macular degeneration J. Ophthalmol. 2018 8125267
[63] Lauermann J L, Eter N and Alten F 2018 Optical coherence tomography angiography offers
new insights into choriocapillaris perfusion Ophthalmologica. 239 74–84
[64] Moult E M, Waheed N K and Novais E A et al 2016 Swept-source optical coherence
tomography angiography reveals choriocapillaris alterations in eyes with nascent geographic
atrophy and drusen-associated geographic atrophy Retina 36Suppl 1 S2–S11
[65] Toto L, Borrelli E and Di Antonio L et al 2016 Retinal vascular plexuses’ changes in dry age-
related macular degeneration, evaluated by means of optical coherence tomography
angiography Retina 36 1566–72
[66] Yannuzzi L A 2010 The Retinal Atlas 1st edn (Edinburgh: Elsevier Saunders)
[67] Inoue M, Jung J J and Balaratnasingam C et al 2016 A comparison between optical
coherence tomography angiography and fluorescein angiography for the imaging of type 1
neovascularization Invest. Ophthalmol. Vis. Sci. 57 OCT314–23
[68] Costanzo E, Miere A and Querques G et al 2016 Type 1 choroidal neovascularization lesion
size: Indocyanine green angiography versus optical coherence tomography Invest. Ophthalmol.
Vis. Sci. 57 OCT307–13
[69] Ryan S J 2013 Retina 5th edn (London: Elsevier)
[70] Dingerkus V L S, Munk M R and Brinkmann M P et al 2019 Optical coherence tomography
angiography (OCTA) as a new diagnostic tool in uveitis J. Ophthalmic Inflamm. Infect 9 10
[71] Khan H A, Iqbal F and Shahzad M A et al 2019 Textural properties of choriocapillaris on
OCTA in healed inflammatory choriocapillaropathies Ophthalmic Surg. Lasers Imaging
Retina 50 566–72
[72] Pichi F, Invernizzi A and Tucker W R et al 2020 Optical coherence tomography diagnostic
signs in posterior uveitis Prog. Retin. Eye Res. 75 100797
[73] Pichi F, Sarraf D and Arepalli S et al 2017 The application of optical coherence tomography
angiography in uveitis and inflammatory eye diseases Prog. Retin. Eye Res. 59 178–201

1-24
 IOP Publishing

Photo Acoustic and Optical Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
Ayman El-Baz and Jasjit S Suri

Chapter 2
Longitudinal optical coherence tomography and
angiography of hyaloid vascular regression in
developing mouse eyes
Tae-Hoon Kim and Xincheng Yao

The hyaloid vascular system transiently nourishes ocular components in the

embryonic eye. Failure of natural regression of the hyaloid vessels results in
congenital ocular disorders. However, the exact mechanism of hyaloid vascular
regression and its correlation with eye development are not well understood.
Previous studies typically relied on end-point examinations with biochemically fixed
tissues, lacking a full description of natural dynamics of the hyaloid vessels along
with eye development. Non-invasive in vivo monitoring is thus required for better
understanding of the dynamic relationships between the hyaloid vessels and the
surrounding ocular tissues. In this chapter, we demonstrate the feasibility of in vivo
longitudinal optical coherence tomography (OCT) and OCT angiography (OCTA)
monitoring of the hyaloid vessels in the developing mouse eyes. OCT allows
structural monitoring of the vessels, and OCTA enables functional assessment of
the vessels. In addition, we will report comparative OCT and OCTA monitoring of
hyaloid vascular regression in wild-type and retinal degeneration 10 (rd10) mice.
Longitudinal OCTA monitoring revealed accelerated regression of the hyaloid
vessels correlated with retinal degeneration in rd10, confirming the close relationship
between the hyaloid vascular system and retinal neurovascular development.
Further OCT/OCTA study would provide in-depth understanding of the complex
mechanism of hyaloid vascular regression in normal and diseased eyes, which is not
only important for advanced study of the nature of the visual system but can also
provide insights into the development of better treatment protocols of congenital
ocular disorders.

doi:10.1088/978-0-7503-2060-3ch2 2-1 ª IOP Publishing Ltd 2021

 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

2.1 Introduction
The hyaloid vessels are the transient sources of nutrients for immature intraocular
tissues in the embryonic eye [1]. Four vascular branches form an entire circulatory
system called the hyaloid vascular system: (1) the hyaloid artery (HA) that arises
from the dorsal ophthalmic artery and enters the eyecup through the embryonic
fissures and divides into (2) the vasa hyaloidea propria (VHP), capillary branches
extending to the posterior surface of the lens. The VHP further divides to form
(3) the tunica vasculosa lentis (TVL), a dense capillary network enclosing the
posterior and lateral surfaces of the lens; (4) the pupillary membrane (PM), which is
an anastomosis of the TVL and the anterior ciliary arteries, covers the anterior
surface of the lens [2, 3]. The HA is a typical arteriole consisting of smooth muscle
cells and endothelial cells connected by tight junctions. The VHP, TVL, and PM are
capillaries consisting of a single layer of non-fenestrated endothelial cells with a
continuous basement membrane and an incomplete layer of pericytes [4]. The
hyaloid vessels are all arteries; thus, blood runs from the HA to the annular vessel
and drains into the choroid vein [5].
The hyaloid vessels are programmed to naturally regress in the human fetus by
34 weeks of gestation, coinciding with development of the retinal vasculature [6].
Failure of apoptosis of the hyaloid vessels results in a congenital ocular anomaly
called persistent fetal vasculature (PFV) that leads to retinal detachment, cataract,
and intraocular hemorrhage [7–9]. Usually, 95% of PFV cases are unilateral, and
few bilateral cases have been reported in conjunction with congenital syndromes
such as Norrie disease and Walker–Warburg syndrome [10]. A study on childhood
blindness and visual loss showed that PFV accounts for about 5% of blindness in
childhood in the United States [11]. In addition, 95% of premature infants and 3% of
full-term infants have persistent hyaloid vessels which can cause multiple ocular
disorders [12, 13]. There are more than 450 000 preterm deliveries each year in the
United States [14], accounting for more than 10% of the 3.9 million newborns
annually [15]. The exact mechanism of the PFV is still unknown as most PFV cases
in humans are sporadic [16, 17]. Only few reports were published on inherited forms
of PFV, and there are no candidate genes identified in humans [10]. Advanced
knowledge of the regression mechanism is thus necessary for a better understanding
of developmental ocular disorders as well as for a better assessment of therapeutic
interventions to prevent vision loss of premature infants due to the PFV.
As experiments on the living fetus are almost impossible, there have been so far
no studies on the mechanism of the postnatal spontaneous regression of the hyaloid
vessels in humans [18]. Instead, experimental studies on the hyaloid vessels are
typically performed with neonatal mice [19–21]. The highly vascularized mouse eye
offers several advantages. In contrast to humans, the mouse pups have persistent
hyaloid vessels at birth, which are known to regress mostly at around postnatal day
16 (P16) [22, 23]. In addition, ocular vascular systems can be genetically manipu-
lated. Several gene mutations have been shown to exhibit the clinical features of the
PFV in mouse models [10]. Moreover, the prenatal development of the mouse eye is
very similar to that of the human eye, and the gestational period of the mouse is

2-2
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

rather short, which makes the mouse an efficient experimental model for the study of
early development of the mammalian eye in line with the hyaloid vascular regression
[24, 25]. However, experimental study using mouse eyes has relied on conventional
fixed-tissue analysis. Since histological examination can only provide snapshots of
vessel remodeling, a dynamic description of vessel remodeling correlated with eye
development can hardly be appreciated. Another caveat of histology is that
postmortem eyeballs can be significantly altered in shape within a few minutes after
extraction [26]. These issues limit the opportunity to investigate coherent relation-
ships between the hyaloid vessels and the surrounding ocular tissues such as the
retina, the lens, and the vitreous. The ability to monitor biological processes in
natural context is important to advance in-depth knowledge of the underlying
mechanism of hyaloid regression.
Thus, in vivo imaging techniques have been extensively explored. Scanning laser
confocal microscopy has visualized intraocular vasculatures including the hyaloid
vessels [23]. High frequency ultrasound microscopy revealed a progressive decrease
in blood flow velocity in the hyaloid vessels with age [27]. Micro-computed
tomography with microvascular corrosion casting technique demonstrated three-
dimensional (3D) imaging of the hyaloid vessels in neonatal mice [28]. Optical
coherence tomography (OCT) successfully visualized parts of the hyaloid vessels in
mouse embryos [29]. With advancement in ophthalmic imaging techniques, fundus
photography, fluorescein angiography, Doppler ultrasound, and OCT have been
used to examine the persistent HA in clinics as well [30–34]. In a recent case report,
OCT angiography (OCTA) was for the first time used to check the presence of blood
flow of the persistent HA [35]. As a new modality of OCT, OCTA provides a label-
free method to visualize the 3D details of ocular vasculatures at high resolution.
OCTA has been widely used for constructing a detailed view of the retinal and
choroidal vasculature both in human and animal eyes [36–42]. Recently, we have
demonstrated, for the first time, the feasibility of in vivo OCT/OCTA monitoring of
hyaloid vascular regression in developing mouse eyes. OCT enabled structural
imaging of the hyaloid vessels, while OCTA allowed functional assessment of the
hyaloid vessels with active blood flow [43–45]. In addition, non-invasive in vivo
OCT/OCTA imaging can not only reduce litter size but also effectively follow up
phenotypic changes, allowing studies of the progression of ocular diseases and the
prenatal effects of pharmacological agents.
In this chapter, we will describe OCT/OCTA imaging of the hyaloid vessels in the
mouse eye and demonstrate immature hyaloid vascular regression due to retinal
degeneration during ocular development.

2.2 Methods
2.2.1 OCT/OCTA imaging system
A custom-designed OCT system was developed for in vivo hyaloid vessel imaging
(figure 2.1). A superluminescent diode (SLD) with a central wavelength λ of 850 nm
and a bandwidth Δλ of 100 nm (D840, Superlum, Cork, Ireland) was used as the
OCT light source, providing 3 μm axial resolution. The lateral resolution was

2-3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 2.1. Optical diagram of the OCT setup. CL: collimation lens; L1, L2, L3: lens; PC: polarization
controller; SLD: superluminescent diode (λ = 850 nm). (Reprinted from [44] with permission from SAGE.)

estimated at 12 μm. The light from the SLD was divided into the sample and
reference arms by a fiber coupler with a splitting ratio of 75:25. The reference arm
had a glass plate for balancing dispersion between the sample and reference paths,
and the sample arm had two scanning mirrors to allow two-dimensional (2D)
scanning. A custom-designed spectrometer was used for OCT recording. A linear
CCD camera (AViiVA EM4; e2v Technologies, Chelmsford, UK) used in the
spectrometer provided a A-scan rate up to 70 kHz. The incident light power on the
mouse cornea was set to 0.95 mW. Technical details are also reported in our
previous publication [46].

2.2.2 OCT/OCTA image acquisition

3D OCT/OCTA volumes were acquired over 1.2 mm (width) × 1.2 mm (length) ×
1.4 mm (height), containing parts of the posterior surface of the lens, the vitreous,
the hyaloid vessels, and the retina (figure 2.2). The optic nerve head (ONH), where
the HA extends from, was located at the center of the imaging area. Four
consecutive OCT B-scans were obtained at each imaging position to compute
speckle variances (SV) for OCTA construction [47]. Thus, each OCTA volume
consisted of a total of 4 × 500 × 500 A-scans.

2.2.3 Quantitative image analysis

Hyaloid vessel density, floater density, retinal thickness, vitreous chamber (VC)
length, and vitreoretinal (VR) length were comprehensively analyzed to see the

2-4
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 2.2. (A) Schematic diagram of the hyaloid vascular system (HVS) in the mouse eye. The green box
indicates the OCT/OCTA imaging area. Representative OCT (B1) and OCTA (B2). In this imaging field, the
hyaloid vessels can be clearly captured together with neighboring ocular tissues including the lens, the retina,
and the vitreous with floaters. (C) Axial length measurements of the posterior segment. Vitreoretinal (VR)
length (white line), vitreous chamber (VC) length (green line) and retinal thickness (yellow line) were separately
measured. (D) Image processing diagram. Constructed en face OCT (D1) and OCTA (D4) images for
morphological and functional vessel density analysis, respectively. Applied image processing ultimately
binarizes hyaloid vasculatures (D2), floaters (D3), and functional hyaloid vessels (D5) for density measure-
ments. (Reprinted from [43], copyright (2019) with permission from Springer (CC BY 4.0).)

relationship between the hyaloid vessels and the surrounding ocular tissues during
eye development. Retinal thickness (from the retinal nerve fiber layer (RNFL) to the
retinal pigment epithelium (RPE)), VC axial length (from the posterior pole of the
lens to the RNFL), and VR axial length (from the posterior pole of the lens to the
RPE) were separately measured, as shown in figure 2.2C. The hyaloid vessel density
was separately analyzed into two categories: (1) ‘structural’ vessel density which is
processed with the OCT dataset and (2) ‘functional’ vessel density which is processed
with the OCTA dataset. Floaters are mainly macrophages and vessel fragments in
the vitreous. Figure 2.2D illustrates the image processing for the vessel density and
floater density analysis.
For the structural vessel density, four repeated OCT B-scans at each scanning
position were first averaged, and an OCT B-scan volume was sliced parallel to the
B-scan cross-section to create a C-scan (en face) stack. Contrast level was adjusted to
individual C-scan images. Next, maximum intensity projection was implemented
in the C-scans (figure 2.2D1). During the projection process, the retinal vasculature
was not included. The Otsu’s automatic thresholding method was then applied
for image binarization [48], followed by floater removal (figure 2.2D2). Binarized
floater images were also produced following a similar procedure to the vessel image
processing, but before the binarization, the vasculature was first removed, followed
by the Otsu’s automatic thresholding (figure 2.2D3).
For the functional vessel density, OCTA volumetric images constructed by
calculating SV from four repeated OCT B-scans were sliced parallel to the B-scan

2-5
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

cross-section to create a C-scan stack. Then the maximum intensity projection was
implemented in the OCTA C-scans (figure 2.2D4). Next, binarization processing
was implemented by using the Otsu’s automatic thresholding method (figure 2.2D5).
Fiji software (http://fiji.sc/Fiji) was used for all the image processing. The vessel
density was defined as the percentage of pixel area occupied by the vasculature in the
binarized image.
All data are expressed as the mean ± standard deviation (SD). Unpaired t-tests
between the animal groups were performed using either Student’s t-test (equal
variances) or Welch’s t-test (unequal variances), and a P-value <0.05 was considered
statistically significant.

2.2.4 Animals
Wild-type mice (WT: C57BL/6J, The Jackson Laboratory, Bar Harbor, ME) and
retinal degeneration 10 mice (Rd10: homozygous for the Pde6brd10 on C57BL/6J
background, The Jackson Laboratory, Bar Harbor, ME) were used in this study.
Rd10 carries a spontaneous point mutation in PDE6B gene (cGMP phosphodies-
terase 6B, rod photoreceptor), and progressive rod photoreceptor death begins at
around P16 at which hyaloid vascular regression is still in progress [49]. Thus, rd10
model provides a unique time window to study correlations of hyaloid vascular
regression with ocular tissue deformities during eye development.
OCT/OCTA measurements were longitudinally conducted. The imaging was
conducted in a room under ambient light condition. Anesthesia was induced
intraperitoneally by injecting a mixture of 100 mg kg−1 ketamine and 5 mg kg−1
xylazine. A drop of ophthalmic mydriatic (1% atropine sulfate ophthalmic solution,
Akorn, Lake Forest, IL) was applied to the imaging eye. The mouse was then placed
in an animal holder integrated with a bite bar and ear bars to minimize motion
artefacts. A cover glass (12–545–80; Microscope cover glass, Thermo Fisher
Scientific, Waltham, MA) along with a drop of eye gel (Severe; GenTeal,
Novartis, Basel, Switzerland) was placed on the imaging eye. After 10 min for
acclimatization and full pupil dilation, the mouse head was fixed for OCT/OCTA
recording. During the experiment, a heating pad was wrapped around the animal
holder to maintain body temperature.
All animal care and experiments were performed in accordance with the
Association for Research in Vision and Ophthalmology (ARVO) statement for
the use of animals in ophthalmic and vision research. All experiments were
performed following the protocols approved by the Animal Care Committee at
the University of Illinois at Chicago.

2.3 Results
2.3.1 OCTA monitoring of hyaloid vascular regression
In this section, we will demonstrate the feasibility of non-invasive in vivo OCT/
OCTA monitoring of the hyaloid vessels in developing mouse eye.
OCT/OCTA images in figure 2.3 show three branches of the hyaloid vascular
system, that is, HA, VHP, and TVL. Multiple capillaries of VHP and vessel fragments

2-6
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 2.3. Representative OCT (A) and OCTA (B) images of a wild-type mouse eye at P14. Ventral (C1 and
D1), temporal/dorsal (C2 and D2), and nasal/ventral (C3 and D3) views of the OCT (C) and OCTA
(D) images. Scale bars: 100 μm. (Reprinted from [44] with permission from SAGE.)

Figure 2.4. Comparative OCT (A) and OCTA (B) images from a wild-type mouse eye measured at P14, P21,
and P28. A1, A2, and A3 illustrate side views of OCT images and A4, A5, and A6 illustrate en face OCT
projection images at the three time points. B1, B2, and B3 illustrate side views of OCTA images and B4, B5,
and B6 illustrate en face OCTA projection images at the three time points. Yellow arrows indicate capillaries in
functional loss. C1, C2, and C3 show floater images, which isolated from A4, A5, and A6, respectively.
(D) Hyaloid vessel density and floater density measurements. Scale bars: 200 μm. (Reprinted from [44] with
permission from SAGE.)

were clearly observed in the OCT image (figure 2.3A), but not in the OCTA image
(figure 2.3B), indicating a lack of blood flow in the VHP. Views from different angles
further confirmed this observation (figures 2.3(C) and (D)). Since the VHP supplies
oxygen to the inner retina until the formation of retinal vasculature [50], it is supposed
to regress by around P14 [22, 27] at the time retinal vasculatures are being formed [6].
Figure 2.4 shows representative OCT/OCTA images longitudinally recorded at
P14, P21, and P28. Comparative C-scan images of OCT (figures 2.4A4–A6) and
OCTA (figures 2.4B4–B6) disclose functionally impaired hyaloid vessels (yellow
arrows). A sequence of regression process also observed that atrophy of the VHP

2-7
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

begins first, followed by regression of capillaries of the TVL, and finally the HA. In
addition, in OCT images, many floaters were observed, and their numbers decreased
with age (figure 2.4C). The floaters are mostly retrolental cell mass and macrophages
[21] and cleaned by phagocytosis [51]. Thus, the vitreous became clear with age.
For quantitative analysis, functional vessel density and floater density were
computed based on OCTA and OCT images, respectively, longitudinally acquired
at P14, P21, and P28 of WT mice (figure 2.4D). The result shows that the functional
vessel density gradually decreased with age (2.72 ± 0.2% [P14] to 0.89 ± 0.1% [P28]),
and the floater density also steadily decreased with age (1.97 ± 0.2% [P14] to 0.62 ±
0.2% [P28]). It is observed that a later phase of regression (P21 to P28) proceeds
slower than a fast regression period (P14 to P21).

2.3.2 Accelerated regression of hyaloid vessels in retinal degeneration mice

In this section, we will demonstrate comparative OCT/OCTA monitoring of hyaloid
vascular regression in WT and rd10 mice. With the onset of photoreceptor
degeneration in rd10, a significant retinal capillary dysfunction was observed [38].
As disturbed retinal development can accompany the persistent hyaloid vessels [7], it
was expected that rd10 mice would reveal a different regression pattern compared to
the healthy control group. Three-dimensional OCT/OCTA images of the hyaloid
vessels were longitudinally acquired in WT and rd10 mice, and the axial length of
VC and VR as well as the retinal thickness were measured to examine a potential
correlation with hyaloid vascular regression (figure 2.2C).
First, the body weights of WT and rd10 mice were compared as the growing body
size is highly correlated with the growth rate of the eye in juvenile animals [52, 53],
and no significant difference was found between two groups (figure 2.5A). There was
also no significant difference in the retinal thickness between two groups at P14.
However, rd10 mice showed a significant reduction of retinal thickness from P17
(figure 2.5B) due to progressive photoreceptor degeneration [38, 54]. It was also
observed that, in conjunction with retinal degeneration, the axial length of the
vitreous abnormally and rapidly changed in rd10. At P14, the VC length was similar
between WT and rd10 mice; however, the VC length of P17 rd10 mice significantly
increased along with the onset of photoreceptor cell death (Rd10: 797.3 ± 38 μm and
WT: 726.3 ± 24.7 μm, P = 0.0014). This result indicates that the inner retinal layers
settled down on the outer retina due to the photoreceptor layer collapsing. After
P17, rd10 mice narrowed the gap of the VC as the VR length rapidly decreased with
age (figures 2.5C and 5D). The VR length of rd10 mice became much shorter than
that of WT mice at P28 (P = 0.0036). These abrupt changes of the vitreoretinal
geometry demonstrate an active emmetropization in rd10 mice, adjusting the axial
length of the eyeball to the focal plane of the ocular optics [55].
Figure 2.6 shows representative OCT/OCTA images of the hyaloid vessels from a
WT mouse and a rd10 mouse. OCTA clearly visualizes the contrast enhanced
vasculatures (figures 2.6B and E). Binarized C-scans of structural OCT vessel image
and functional OCTA vessel image were prepared for quantitative analysis.
Composite images (figures 2.6C and F) represent the binarized structural vessels

2-8
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 2.5. (A) Body weight measurement. (B) Retinal thickness measurement. Photoreceptor degeneration
resulted in significant thinning of the retinal thickness in rd10 mice. (C) VC length measurement. Onset of
photoreceptor degeneration induced rapid extension of the VC length (724.0 ± 16.3 μm [P14] to 797.3 ±
38.0 μm [P17]). Subsequently, the VC length of rd10 mice steadily decreased with aging: 784.3 ± 27.5 μm (P21)
and 753.7 ± 14.1 μm (P28). WT showed little change in the VC length throughout the experimental period:
724.3 ± 17.7 μm (P14) and 716.2 ± 14.7 μm (P28). (D) VR length measurement. Rd10 mice initially had a wide
VR length in response to the onset of photoreceptor degeneration, followed by a sharp narrowing phase. The
VR length in WT mice showed a slight declining trend (921.2 ± 17.3 μm [P14] to 903.4 ± 7.2 μm [P28]). N = 7
for WT and N = 7 for rd10. P values for an unpaired t-test between WT and rd10 mice are indicated by
asterisks: *P < 0.05. (Reprinted from [43], copyright (2019) with permission from Springer (CC BY 4.0).)

(red color), floaters (white color), and functional vessels (green color). It is readily
observed that vascular branches and floaters decreased with age in two mouse
groups. Although the structural vessel density (figure 2.6G) and floater density
(figure 2.6I) were similar between the two groups until P24, rd10 mice showed a
rapid vascular dropout at P28 (1.96 ± 0.9%) (figure 2.6G). Notably, almost complete
functional loss was found in rd10 mice at P24 before obliteration of the hyaloid
vasculature (figure 2.6H). While, WT mice revealed a gradual declining trajectory in
vessel density with age, and some vessels were still functional even at P28 (4.11 ±
0.5%). In rd10 mice, the floater density abruptly increased at P28 due to the
accelerated hyaloid vascular regression (figure 2.6I). These results demonstrate that
retinal degeneration during ocular development affects the hyaloid vascular regres-
sion process.

2-9
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 2.6. Representative OCT (A, D) and OCTA (B, E) images from a WT mouse and a rd10 mouse,
longitudinally recorded at P14, P17, P21, P24, and P28. In binarized en face composite images (C) and (F), red,
white, and green colors represent the hyaloid vasculature, floaters, and functional vessels, respectively. The
retinal and choroidal vasculature was excluded from en face image construction. Yellow arrows indicate
capillaries in functional loss, that is, a lack of blood flow. Scale bars: 200 μm. (G) Hyaloid vasculature density
measurements from OCT datasets. Accelerated regression of hyaloid vessel occurred in rd10 mice compared to
WT mice (4.11 ± 0.5% [WT at P28] and 1.98 ± 0.9% [Rd10 at P28], P = 0.0002). (H) Functional hyaloid vessel
density measurements from the OCTA datasets. Functional OCTA revealed even earlier physiological
dysfunction in rd10 mice before obliteration of the hyaloid vasculature (2.06 ± 0.6% [WT at P24] and 1.05
± 0.5% [Rd10 at P24], P = 0.0061). (I) Floater density measurements. Floater density steadily decreased in WT
with age. N = 7 for WT and N = 7 for rd10. P values for an unpaired t-test between WT and rd10 mice are
indicated by asterisks: *P < 0.05. (Reprinted from [43], copyright (2019) with permission from Springer (CC
BY 4.0).)

2.4 Discussion
In this chapter, we introduced OCT/OCTA imaging of the hyaloid vessels in the
developing mouse eyes. OCT imaging allowed morphological monitoring of the
hyaloid vessels and surrounding tissues. OCTA imaging enabled functional assess-
ment of the hyaloid vessels, that is, insufficient or lacking blood flow, before
morphological changes occur. Accelerated hyaloid vascular regression was also
observed in rd10 mice, following the onset of rod cell degeneration. This premature
regression is a rare case as most mutant mouse models have shown persistent hyaloid
vessels in the case of malformation of retinal vasculatures [56]. On the other hand, all
WT mice had functional hyaloid vessels over P28, which is consistent with a
previous fluorescein angiographic study where the hyaloid vessels were found in all
WT mice at 8–25 weeks old [57]. Since the primary function of the hyaloid vessels is
to nurture the developing ocular components, the sustaining functional vessels in
WT mice may be still involved in later developmental phases of the eye [58–60].

2-10
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Accordingly, the premature regression of the hyaloid vessels in rd10 mice could
impinge on the normal eye development. In fact, rd10 mice have shown shorter axial
length of the eyeball and lighter eye weight compared to WT mice [61, 62].
The exact mechanism of hyaloid vascular regression under the normal condition
is not fully understood yet. Multiple contributing factors are suggested to be directly
involved in the regression process. Meeson et al proposed two stages in regression.
The first stage is the apoptosis of a single endothelial cell, which restricts the
capillary lumen and subsequently imposes either a temporary or permanent block to
blood flow. In the second stage, the cessation of blood flow consequently causes
synchronous apoptosis of downstream endothelial cells in the affected segment [63].
Reduction of shear stress on the endothelial cells due to reduced blood flow can also
stimulate expression of vasoconstrictors [64], and vasoconstriction of the hyaloid
vessels appears in the early phase of regression process [65]. Mitchell et al
demonstrated widespread apoptosis of the endothelial cells in the TVL, and the
vascular regression was associated with loss of capillary integrity and phagocytosis
of the apoptotic endothelium by macrophages [66]. Macrophages contribute
significantly in hyaloid vascular regression involving the induction of apoptosis,
the occlusion, and obliteration of atrophic vessels [3, 19–21]. Kishimoto et al further
confirmed a narrowed vessel width during regression of the VHP and found severe
vascular congestions due to aggregated neutrophilic leukocytes together with
erythrocytes in the lumen of hyaloid vessels [21]. The interaction between macro-
phages and blood vessels was found to involve Wnt7b signaling derived from
macrophages via FZD4 receptor and LRP5 co-receptor on endothelial cells. As a
result, mice with FZD4 or LRP5 null mutations or a hypomorphic Wnt7b allele
were accompanied by the disruption of hyaloid vascular regression [20, 67, 68]. In
our study, cessation of blood flow was clearly demonstrated in the OCTA images
(figures 2.4 and 2.6). In addition, fragmented threadlike hyaloid vessels were
frequently observed in OCT images, which may indicate the presence of aggregated
blood cells in the lumen (figure 2.3). This result supports that the reduced blood flow
by vascular occlusion with blood cells is the major triggering source of the hyaloid
vascular regression under the natural context.
The hyaloid vascular regression is also considerably affected by retinal oxygen
levels that regulate the expression of vascular endothelial growth factor (VEGF), a
signal protein related to the formation of the hyaloid vessels as well as its
maintenance [69]. It was reported that low oxygenation in the retina due to
malformation of retinal vasculatures induced upregulation of VEGF levels, sub-
sequently resulting in persistent hyaloid vessels [70]. The bilateral PFV occurs in
humans with Norrie disease [71], and mice lacking Norrin showed upregulation of
VEGF levels in neurons, and reducing VEGF levels corrected the vascular
abnormalities in mice [68, 72]. In contrast, high oxygenation can lead to down-
regulation of VEGF levels that can cause premature regression of the hyaloid vessels
[73]. It was confirmed that inactivation of VEGF accelerated hyaloid vascular
regression [69]. A study demonstrated that melanopsin, a light-sensitive molecule
found in a small subset of retinal ganglion cells, responds to light exposure and
restrains VEGF-A expression, which consequently allows hyaloid vascular

2-11
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

regression to occur [74]. In our study, rd10 mice manifested accelerated hyaloid
vascular regression (figure 2.6), and it is plausible to speculate that the retinal
hyperoxia in rd10 mice may influence the regression process. In fact, significant
attenuation of retinal blood flow in the deep capillary plexus occurred simulta-
neously with rod photoreceptor apoptosis [38], which enhances oxygen diffusion
towards the vitreous due to a lack of autoregulating capacity in the choroid
circulation [75]. Another intriguing factor to consider is neurotransmitter dopamine
released by dopaminergic amacrine cells (DACs) in the retina, which can directly
inhibit VEGF receptor-2 signaling in hyaloid vascular endothelial cells and
consequently accelerate hyaloid vascular regression [56]. The inner retinal neurons
in rd10 mice undergo extensive remodeling after loss of rod photoreceptors, which
makes DACs exhibiting aberrant activity leading to impaired dopamine metabolism
[76]. It was found that the dopamine levels in rd10 mouse retinas were significantly
higher than those in WT mouse retinas [62]. Thus, the abnormal dopamine levels
due to the inner retinal remodeling can be another triggering source of the
accelerated hyaloid regression in rd10 mice. Overall, the rd10 study suggests an
unknown signaling pathway between the retinal and hyaloid vessels [5]. It has been
noted that the disturbed formation of the deep capillary plexus in mutant mouse
retinas directly affects hyaloid vessel regression [7].

2.5 Conclusion
In summary, we have demonstrated the feasibility of longitudinal in vivo OCT/
OCTA monitoring of the hyaloid vascular regression in developing mouse eyes.
Comparative OCT and OCTA revealed accelerated regression of the hyaloid vessels
in rd10 mice due to rod photoreceptor degeneration during ocular development,
which clearly indicates a functional link between the retinal neurovascular system
and the hyaloid vascular system. We expect that further longitudinal OCT/OCTA
investigation of the retinal and hyaloid vessels in normal and mutant mouse models
will be valuable to unravel the unknown mechanisms of the PFV. Label-free OCT/
OCTA will also provide an imaging platform for longitudinal monitoring of
therapeutic treatments of the animal models, fostering drug developments for
effective treatment of PFV.

References
[1] Usui Y, Westenskow P D, Murinello S, Dorrell M I, Scheppke L, Bucher F, Sakimoto S,
Paris L P, Aguilar E and Friedlander M 2015 Angiogenesis and eye disease Annu. Rev. Vis.
Sci. 1 155–84
[2] Yee K M P, Feener E P, Gao B, Aiello L P, Madigan M C, Provis J, Ross-Cisneros F N,
Sadun A A and Sebag J I D 2014 Vitreous cytokines and regression of the fetal hyaloid
vasculature ed J Sebag Vitreous: in Health and Disease (New York: Springer) 41–55
[3] Zhang H, Tse J, Hu X, Witte M, Bernas M, Kang J, Tilahun F, Hong Y K, Qiu M and Chen
L 2010 Novel discovery of LYVE-1 expression in the hyaloid vascular system Investig.
Ophthalmol. Vis. Sci. 51 6157–61
[4] Sebag J 2015 Vitreous: In Health and Disease (Berlin: Springer)

2-12
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[5] Saint-Geniez M and D’Amore P A 2004 Development and pathology of the hyaloid,
choroidal and retinal vasculature Int. J. Dev. Biol. 48 1045–58
[6] Stahl A et al 2010 The mouse retina as an angiogenesis model Investig. Ophthalmol. Vis. Sci.
51 2813–26
[7] Fruttiger M 2007 Development of the retinal vasculature Angiogenesis. 10 77–88
[8] Silbert M and Gurwood A S 2000 Persistent hyperplastic primary vitreous Clin. Eye Vis.
Care 12 131–7
[9] Goldberg M F 1997 Persistent fetal vasculature (PFV): an integrated interpretation of signs
and symptoms associated with persistent hyperplastic primary vitreous (PHPV). LIV
Edward Jackson Memorial Lecture Am. J. Ophthalmol. 124 587–626
[10] Chen C, Xiao H and Ding X 2019 Persistent fetal vasculature The Asia-Pacific J.
Ophthalmol. 8 86–95
[11] Mets M B 1999 Childhood blindness and visual loss: an assessment at two institutions
including a ‘new’ cause Trans. Amer. Ophthal. Soc 97 653–96
[12] Fineman M S and Ho A C 2012 Color Atlas and Synopsis of Clinical Ophthalmology:
Retina 2nd edn (Baltimore, MD: Wolters Kluwer/Lippincott Williams & Wilkins)
[13] Hartnett M E, Trese M and Capone A 2015 Pediatric Retina: Medical and Surgical
Approaches (Philadelphia, PA: Wolters Kluwer)
[14] Liegl R, Hellstrom A and Smith L E 2016 Retinopathy of prematurity: the need for
prevention Eye Brain 8 91–102
[15] Facts About Retinopathy of Prematurity (ROP) https://neinihgov/health/rop/rop
[16] Vasavada A R, Vasavada S A, Bobrova N, Praveen M R, Shah S K, Vasavada V A, Pardo
A J, Raj S M and Trivedi R H 2012 Outcomes of pediatric cataract surgery in anterior
persistent fetal vasculature J. Cataract Refract. Surg. 38 849–57
[17] Khaliq S, Hameed A, Ismail M, Anwar K, Leroy B, Payne A M, Bhattacharya S S and
Mehdi S Q 2001 Locus for autosomal recessive nonsyndromic persistent hyperplastic
primary vitreous Investig. Ophthalmol. Vis. Sci. 42 2225–8
[18] Li J, Zhang J and Lu P 2019 Regression of fetal vasculature and visual improvement in
nonsurgical persistent hyperplastic primary vitreous: a case report BMC Ophthalmol. 19 161
[19] Lang R A and Bishop J M 1993 Macrophages are required for cell death and tissue
remodeling in the developing mouse eye Cell. 74 453–62
[20] Lobov I B et al 2005 WNT7b mediates macrophage-induced programmed cell death in
patterning of the vasculature Nature 437 417–21
[21] Kishimoto A, Kimura S, Nio-Kobayashi J, Takahashi-Iwanaga H, Park A M and Iwanaga
T 2018 Histochemical characteristics of regressing vessels in the hyaloid vascular system of
neonatal mice: novel implication for vascular atrophy Exp. Eye Res. 172 1–9
[22] Ito M and Yoshioka M 1999 Regression of the hyaloid vessels and pupillary membrane of
the mouse Anat. Embryol. 200 403–11
[23] Ritter M R, Aguilar E, Banin E, Scheppke L, Uusitalo-Jarvinen H and Friedlander M 2005
Three-dimensional in vivo imaging of the mouse intraocular vasculature during development
and disease Investig. Ophthalmol. Vis. Sci. 46 3021–6
[24] Pei Y F and Rhodin J A 1970 The prenatal development of the mouse eye Anatomical Rec.
168 105–25
[25] Pardue M T, Stone R A and Iuvone P M 2013 Investigating mechanisms of myopia in mice
Exp. Eye Res. 114 96–105

2-13
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[26] Jiang X, Kurihara T, Kunimi H, Miyauchi M, Ikeda S I, Mori K, Tsubota K, Torii H and
Tsubota K 2018 A highly efficient murine model of experimental myopia Sci. Rep. 8 2026
[27] Brown A S, Leamen L, Cucevic V and Foster F S 2005 Quantitation of hemodynamic
function during developmental vascular regression in the mouse eye Investig. Ophthalmol.
Vis. Sci. 46 2231–7
[28] Atwood R C, Lee P D, Konerding M A, Rockett P and Mitchell C A 2010 Quantitation of
microcomputed tomography-imaged ocular microvasculature Microcirculation 17 59–68
[29] Larina I V, Syed S H, Sudheendran N, Overbeek P A, Dickinson M E and Larin K V 2012
Optical coherence tomography for live phenotypic analysis of embryonic ocular structures in
mouse models J. Biomed. Opt. 17 081410–1
[30] Sundararajan M, Jansen M E and Gupta M 2018 Unilateral persistence of the hyaloid artery
causing vitreopapillary and vitreomacular traction JAMA Ophthalmol. 136 e180221
[31] Delaney W V Jr Prepapillary hemorrhage and persistent hyaloid artery Am. J. Ophthalmol.
198090 419–21
[32] Chen T L and Yarng S S 1993 Vitreous hemorrhage from a persistent hyaloid artery Retina
13 148–51
[33] Gonçalves A, Cruysberg J R, Draaijer R W, Sellar P W, Aandekerk A L and Deutman A F
1996 Vitreous haemorrhage and other ocular complications of a persistent hyaloid artery
Doc. Ophthalmol. Adv. Ophthalmol. 92 55–9
[34] Azrak C, Campos-Mollo E, Lledó-Riquelme M, Ibañez F A and Toldos J J 2011 Vitreous
haemorrhage associated with persistent hyaloid artery Archivos de la Sociedad Espanola de
Oftalmologia 86 331–4
[35] Jeon H, Kim J and Kwon S 2019 OCT angiography of persistent hyaloid artery: a case
report BMC Ophthalmol. 19 141
[36] Park J R, Choi W, Hong H K, Kim Y, Jun Park S, Hwang Y, Kim P, Joon Woo S, Hyung
Park K and Oh W Y 2016 Imaging laser-induced choroidal neovascularization in the rodent
retina using optical coherence tomography angiography Investig. Ophthalmol. Vis. Sci. 57
Oct331–40
[37] Liu W, Li H, Shah R S, Shu X, Linsenmeier R A, Fawzi A A and Zhang H F 2015
Simultaneous optical coherence tomography angiography and fluorescein angiography in
rodents with normal retina and laser-induced choroidal neovascularization Opt. Lett. 40
5782–5
[38] Kim T H, Son T, Lu Y, Alam M and Yao X 2018 Comparative optical coherence
tomography angiography of wild-type and rd10 mouse retinas Transl. Vis. Sci. Technol 7 42
[39] Borrelli E, Sarraf D, Freund K B and Sadda S 2018 OCT angiography and evaluation of the
choroid and choroidal vascular disorders Prog. Retin. Eye Res. 67 30–55
[40] Yao X, Alam M N, Le D and Toslak D 2020 Quantitative optical coherence tomography
angiography: a review Exp. Biol. Med. 245 301–12
[41] Jia Y et al 2015 Quantitative optical coherence tomography angiography of vascular
abnormalities in the living human eye Proc. Natl. Acad. Sci. USA 112 E2395–402
[42] Hwang T S, Jia Y, Gao S S, Bailey S T, Lauer A K, Flaxel C J, Wilson D J and Huang D
2015 Optical coherence tomography angiography features of diabetic retinopathy Retina 35
2371–6
[43] Kim T H, Son T, Le D and Yao X 2019 Longitudinal OCT and OCTA monitoring reveals
accelerated regression of hyaloid vessels in retinal degeneration 10 (rd10) mice Sci. Rep. 9
16685

2-14
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[44] Kim T H, Son T and Yao X 2019 Feature article: functional OCT angiography reveals early
physiological dysfunction of hyaloid vasculature in developing mouse eye Exp. Biol. Med.
244 819–23
[45] Kim T-H, Son T and Yao X 2019 Longitudinal optical coherence tomography angiography
of hyaloid vessels in the developing mouse eye Proc. SPIE 10858, Ophthalmic Technologies
XXIX 108581N
[46] Son T, Wang B, Thapa D, Lu Y, Chen Y, Cao D and Yao X 2016 Optical coherence
tomography angiography of stimulus evoked hemodynamic responses in individual retinal
layers Biomed. Opt. Express 7 3151–62
[47] Mahmud M S, Cadotte D W, Vuong B, Sun C, Luk T W, Mariampillai A and Yang V X
2013 Review of speckle and phase variance optical coherence tomography to visualize
microvascular networks J. Biomed. Opt. 18 50901
[48] Otsu N 1979 A threshold selection method from gray-level histograms IEEE Trans. Syst.
Man Cybern. 9 62–6
[49] Chang B et al 2007 Two mouse retinal degenerations caused by missense mutations in the
beta-subunit of rod cGMP phosphodiesterase gene Vis. Res 47 624–33
[50] Lutty G A and McLeod D S 2018 Development of the hyaloid, choroidal and retinal
vasculatures in the fetal human eye Prog. Retin. Eye Res 62 58–76
[51] Vrolyk V, Haruna J and Benoit-Biancamano M O 2018 Neonatal and Juvenile Ocular
development in Sprague-Dawley Rats: a histomorphological and immunohistochemical
study Vet. Pathol. 55 310–30
[52] Prashar A, Hocking P M, Erichsen J T, Fan Q, Saw S M and Guggenheim J A 2009
Common determinants of body size and eye size in chickens from an advanced intercross line
Exp. Eye Res. 89 42–8
[53] Wang D, Ding X, Liu B, Zhang J and He M 2011 Longitudinal changes of axial length and
height are associated and concomitant in children Investig. Ophthalmol. Vis. Sci. 52 7949–53
[54] Lu Y, Kim T-H and Yao X 2020 Comparative study of wild-type and rd10 mice reveals
transient intrinsic optical signal response before phosphodiesterase activation in retinal
photoreceptors Exp. Biol. Med. 245 360–7
[55] Wallman J and Winawer J 2004 Homeostasis of eye growth and the question of myopia
Neuron 43 447–68
[56] Nguyen M T et al 2019 An opsin 5-dopamine pathway mediates light-dependent vascular
development in the eye Nat. Cell Biol. 21 420–9
[57] McLenachan S, Magno A L, Ramos D, Catita J, McMenamin P G, Chen F K, Rakoczy E P
and Ruberte J 2015 Angiography reveals novel features of the retinal vasculature in healthy
and diabetic mice Exp. Eye Res. 138 6–21
[58] Zhou X, Shen M, Xie J, Wang J, Jiang L, Pan M, Qu J and Lu F 2008 The development of the
refractive status and ocular growth in C57BL/6 mice Investig. Ophthalmol. Vis. Sci. 49 5208–14
[59] Tkatchenko T V, Shen Y and Tkatchenko A V 2010 Analysis of postnatal eye development
in the mouse with high-resolution small animal magnetic resonance imaging Investig.
Ophthalmol. Vis. Sci. 51 21–7
[60] Schmucker C and Schaeffel F 2004 A paraxial schematic eye model for the growing C57BL/6
mouse Vis. Res 44 1857–67
[61] Wisard J, Chrenek M A, Wright C, Dalal N, Pardue M T, Boatright J H and Nickerson J M
2010 Non-contact measurement of linear external dimensions of the mouse eye J. Neurosci.
Methods 187 156–66

2-15
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[62] Park H, Tan C C, Faulkner A, Jabbar S B, Schmid G, Abey J, Iuvone P M and Pardue M T
2013 Retinal degeneration increases susceptibility to myopia in mice Mol. Vis. 19 2068–79
[63] Meeson A, Palmer M, Calfon M and Lang R 1996 A relationship between apoptosis and
flow during programmed capillary regression is revealed by vital analysis Development 122
3929–38
[64] Lobov I B et al 2011 The Dll4/Notch pathway controls postangiogenic blood vessel
remodeling and regression by modulating vasoconstriction and blood flow Blood. 117
6728–37
[65] Browning J, Reichelt M E, Gole G A and Massa H 2001 Proximal arterial vasoconstriction
precedes regression of the hyaloid vasculature Curr. Eye Res. 22 405–11
[66] Mitchell C A, Risau W and Drexler H C 1998 Regression of vessels in the tunica vasculosa
lentis is initiated by coordinated endothelial apoptosis: a role for vascular endothelial growth
factor as a survival factor for endothelium Dev. Dyn. 213 322–33
[67] Kato M et al 2002 Cbfa1-independent decrease in osteoblast proliferation, osteopenia, and
persistent embryonic eye vascularization in mice deficient in Lrp5, a Wnt coreceptor J. Cell
Biol. 157 303–14
[68] Xu Q et al 2004 Vascular development in the retina and inner ear: control by Norrin and
Frizzled-4, a high-affinity ligand-receptor pair Cell. 116 883–95
[69] Yoshikawa Y, Yamada T, Tai-Nagara I, Okabe K, Kitagawa Y, Ema M and Kubota Y
2016 Developmental regression of hyaloid vasculature is triggered by neurons J. Exp. Med
213 1175–83
[70] Claxton S and Fruttiger M 2003 Role of arteries in oxygen induced vaso-obliteration Exp.
Eye Res 77 305–11
[71] Ye X, Wang Y and Nathans J 2010 The Norrin/Frizzled4 signaling pathway in retinal
vascular development and disease Trends Mol. Med. 16 417–25
[72] Rattner A, Wang Y, Zhou Y, Williams J and Nathans J 2014 The role of the hypoxia
response in shaping retinal vascular development in the absence of Norrin/Frizzled4
signaling Investig. Ophthalmol. Vis. Sci. 55 8614–25
[73] Penn J S, Li S and Naash M I 2000 Ambient hypoxia reverses retinal vascular attenuation in
a transgenic mouse model of autosomal dominant retinitis pigmentosa Investig. Ophthalmol.
Vis. Sci. 41 4007–13
[74] Rao S, Chun C, Fan J, Kofron J M, Yang M B, Hegde R S, Ferrara N, Copenhagen D R
and Lang R A 2013 A direct and melanopsin-dependent fetal light response regulates mouse
eye development Nature 494 243–6
[75] Yu D Y and Cringle S J 2005 Retinal degeneration and local oxygen metabolism Exp. Eye
Res. 80 745–51
[76] Ivanova E, Yee C W and Sagdullaev B T 2016 Disruption in dopaminergic innervation
during photoreceptor degeneration J. Comp. Neurol. 524 1208–21

2-16
 IOP Publishing

Photo Acoustic and Optical Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
Ayman El-Baz and Jasjit S Suri

Chapter 3
Stripe noise removal and vessel segmentation of
OCTA images
Xiyin Wu, Dongxu Gao, Davide Borroni, Savita Madhusudhan and Yalin Zheng

Optical coherence tomography angiography (OCTA) is a kind of retinal imaging

technology that provides important information on fine retinal vessels in a non-
invasive and non-contact way. The effective analysis of retinal blood vessels is
valuable for the investigation and diagnosis of vascular and vascular-related
diseases, for which accurate vessel segmentation is a vital first step. OCTA images
are always affected by some stripe noises artifacts, which will impede correct
segmentation and should be removed. To address this issue, we present a two-stage
strategy in this chapter, in which stripe noise removal is addressed by image
decomposition and vessel segmentation is solved by an active contours approach.
Besides, a joint model is also given in this chapter to improve the speed of the
algorithm while retaining the quality of the segmentation and destriping. The
experimental results are presented on both simulated and real retinal imaging
data, demonstrating the effective performance of the joint model for segmenting
vessels from the OCTA images corrupted by stripe noise. Furthermore, the stripe
direction influence is tested to prove the flexibility of the joint model.

3.1 Introduction
Retinal blood vessel health is integral to high-quality human vision and provides
useful clinical information. Changes in the retinal vasculature have a close relation-
ship with many ophthalmological and cardiovascular diseases, such as diabetic
retinopathy (DR) and age-related macular degeneration (AMD) [1]. To acquire
imagery of fine retinal vessels, the commonly used techniques are fluorescein
angiography (FA) and indocyanine green angiography (ICGA) [2]. Both FA and
ICGA require intravenous dye injections, which can have adverse side effects
and still only provide information relating to superficial blood vessels. Recently,

doi:10.1088/978-0-7503-2060-3ch3 3-1 ª IOP Publishing Ltd 2021

 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

an innovative technology called optical coherence tomography angiography

(OCTA) has emerged as an effective way of visualizing the retinal vessels up to
the capillary level [3], which previously can only be achieved by invasive and
potentially life-threatening angiographic techniques [4]. OCTA is a functional
extension of OCT imaging, allowing us to visualize the microvasculature of the
eye, without the need for invasive and potentially harmful dyes [3]. Unlike FA and
ICGA, OCTA is a non-invasive, fast, depth-resolved approach. OCTA can fully
visualize choroidal neovascularization in AMD and small retinal neovascularization
in DR, which are difficult to identify with FA. Therefore, the usage of OCTA in the
diagnosis of vascular diseases is expected to increase significantly soon.
The acquisition process of OCTA images is illustrated in figure 3.1. It can be
viewed as a functional extension of well-established optical coherence tomography
(OCT) imaging to visualize the microvasculature of the eye [5]. OCTA reconstructs
blood vessels by detecting moving particles such as red blood cells in the tissue by
deriving phase [6] or amplitude [7] variances from the repeated OCT scans at the
same location. A 3D OCTA map can be produced by taking sequential cross-
sectional OCTA scans of the scanned tissue [8]. For the ease of visualizing the retinal
vessels, projections of the acquired 3D OCTA map into 2D en face images are often
used. As shown in figures 3.1(d) and (e), retinal vessels at a certain depth of the retina
can be generated by selecting certain layers of the retina from the projection. It has
the advantage of acquiring three-dimensional datasets, allowing us to selectively
view each layer of the retinal capillary plexuses and choroidal vasculature. For
example, the vessels of the whole retina (WR), superficial vascular plexus (SVP),

Figure 3.1. The acquisition process of an OCTA image. (a) Infrared image illustrating the scanning position of
OCT. (b) OCT structural volume with repeated scans at each scanning location. (c) OCTA volume data
(yellow dots) superimposed on the OCT image. OCTA is acquired by phase or amplitude variances from
repeated OCT scans at each location. (d) A zoomed-in region of (c). (e) Projection of OCTA map by selecting
certain layers: SVP, DVP, AL and WR.

3-2
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

deep vascular plexus (DVP) and avascular layer (AL) are most often used in
practice:
• Whole retina (WR): from the internal limiting membrane to the Bruch’s
membrane.
• Superficial vascular plexus (SVP): from the internal limiting membrane to the
inner plexiform layer.
• Deep vascular plexus (DVP): from the inner plexiform layer to the outer
plexiform layer.
• Avascular layer (AL): from the outer plexiform layer to the Bruch’s membrane.

Retinal vessel segmentation is the first and most important step to detect the
retinal vasculature. The segmentation of retinal vessels is a valuable precursor for
further processing and analysis, such as retinal image registration, feature extraction
and localization of retinal structures like the fovea and the optic disk [9–12]. As a
relatively new modality, few studies exist analyzing the retinal vessels in OCTA and
the OCTA vessel segmentation is still at an early stage in its development. Eladawi
et al [13] presented a joint Markov–Gibbs random field method to segment blood
vessels from OCTA scans. The authors further estimated three local features from
the segmented vessels to distinguish the status of DR patients [14]. Compared with
OCTA, research in retinal vessel segmentation with color fundus images has a longer
history. Many methods have been proposed over the last two decades, such as active
contour models, wavelets methods, Gaussian mixture models, Adaboost and
support vector machine, to name a few. One of the most commonly used
segmentation approaches is active contours, such as the Chan–Vese model [15].
The Chan–Vese model applied global statistics to the extraction of objects and was
useful for objects with homogeneous intensity. To handle the non-uniform situation
of the Chan–Vese model, Sum et al [16] developed a modified version by combining
the local image contrast into a level set based active contour. Bashir et al [17] grew a
Ribbon of Twins active contour model to locate vessel edges under different
conditions. Zhao et al [18] segmented retinal vessels by developing an infinite
perimeter active counter model with hybrid region information. However, color
fundus images cannot provide depth information, limiting the analysis of choroidal
neovascularization and small retinal neovascularization.
There is a main limitation in the mentioned literature in contributing to clinical
diagnostics. Previous works on OCTA images have neglected the unavoidable noise
problem. Additional image noise can arise from the OCTA image acquisition, eye
motion and image pre-processing strategies. One of the most common types of noise
is white horizontal or vertical stripe noise (see figure 3.1), which results from patient
eye motion [19]. During a single OCTA volumetric scan, which usually lasts several
seconds (depending on the OCT devices, e.g. 3–5 s [20]), the involuntary eye
movement of patients gives rise to motion artifacts, including tremors, micro-
saccades and drift [20]. These eye motions cause intensity distribution shift within
B-scans at each position, and produce visible horizontal stripes in the OCTA images
[21]. These stripes artifacts affect the visualization and interpretation of OCTA
images and even the clinical diagnosis, and need to be tackled before segmenting the

3-3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

retinal vessels to achieve effective, good-quality segmentation. Therefore, commer-

cial OCTA devices adopt some motion correction methods to reduce this problem,
such as Angioplex and Spectralis OCT (Heidelberg Engineering, Heidelberg,
Germany) [22]. A scanning laser ophthalmoscope (SLO) has been used to resume
the scan by adjusting the scan to the right location [23, 24]. Orthogonal registration
combines the information in horizontal and vertical OCT volume scans to produce
one OCTA map [25]. Although motion correction methods adopted in these devices
can improve the scan quality, residual non-homogeneous lines and distorted lines
remain. Therefore, new image processing solutions are required to address this
inherent problem of OCTA.
To overcome this limitation, the stripe noise removal (‘destriping’) technique is
introduced to alleviate this inherent nature of OCTA images during the acquisition
process. Although there is scarce literature about removing stripe noise in OCTA
images, images of other modalities suffer from similar noise and lots of related
works have been explored. These techniques generally based on transform domain
or spatial domain. The stripe noise is considered as specific frequencies in transform
domain based methods. These methods are very simple to implement and are used
widely in lots of imaging modalities [26–28]. However, most of their theory is based
on the assumptions that the stripes are periodic and therefore their performance is
limited in specific cases. The spatial based methods usually employ deep networks or
construct an optimization model. The deep based spatial methods learn the
characters of stripe noise in a supervised way and differentiate the stripe from the
image structure from a global view [29]. The limitation of these methods is that they
take a long time to train models, and the training often requires numerous images as
well as their corresponding ground truth images, which are hard to obtain in real
applications. The optimization based spatial methods consider stripe removal as an
ill-posed inverse problem, in which the stripe free image and stripes are characterized
based on prior information. These methods are making momentous progress in
different applications, such as remote sensing images and microscopy imaging
[30–32]. The destriping models based on optimization theory inspire us to take a
similar approach to OCTA imaging.
In this chapter, we will first present a two-stage strategy for stripe noise removal
and retinal vessel segmentation in OCTA images. This strategy removes stripe noise
and Gaussian noise from the image before segmenting the denoised image. Then a
joint model will be given to achieve the two tasks simultaneously. For stripe noise
removal, the optimization based spatial theory inspires us to take a similar approach
in OCTA imaging. For vessel segmentation, a good choice is active contour models,
which can not only provide smooth and closed contours but also achieve subpixel
accuracy on vessel boundaries.

3.2 Two-stage strategy (TSS)

A two-stage strategy is introduced for segmenting OCTA images affected by stripe
noise. In this strategy, we first remove the stripe noise by an optimization based
model and subsequently segment the retinal vessels with an active contour model.

3-4
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

3.2.1 The first stage: stripe noise removal of OCTA images

3.2.1.1 Image observation formula
Considering the stripe acquired along with the image content, we proceed with an
image decomposition strategy, i.e., the given noisy image is decomposed to the
desired clean image, the stripe noise and small noise. The model of the observed
image can be formulated as:
I=C+S+N (3.1)
where I is the given noisy image; C denotes the desired clean image; S is the additive
stripe component; and N represents the linear assumption error and small noise.
Stripe noise removal aims to estimate both C and S simultaneously from I .

3.2.1.2 Priors of stripe noise and clean OCTA image

As shown in figure 3.1, the stripe noise in OCTA images has a salient structural
characteristic, showing in either the horizontal or vertical direction. To analyze the
characteristic of stripe noise, we constructed an image with horizontal stripes (see
figure 3.2(b)) and vertical stripes (seen figure 3.2(d)). It is found that the rank of
horizontal stripe matrix is equal to 1 and that the rank of the vertical stripe matrix is
also equal to 1. The rank value of the stripe matrix indicates that its spanned
subspace can be expressed as a low-rank constraint, which is expressed as rank(S ).
Moreover, we studied the rank change before and after the OCTA image was
degraded by stripe noise. A 256 × 256 OCTA image (see figure 3.2(a)) is corrupted
by adding the horizontal stripes in figure 3.2(b) (or the vertical stripes in
figure 3.2(d)) and the resulting degraded image is shown in figure 3.2(c) (or
figure 3.2(e)). Their corresponding natural logarithms of singular values are depicted
in figure 3.2(h) (or figure 3.2(j)). While the singular values of both the clean OCTA
image and degraded image approximate to zero at rank 256 slowly, the singular
value of the stripe image approximates to zero at rank 1 rapidly. The trends of clean

Figure 3.2. Low rank prior of stripes. (a) Simulated OCTA image. (b) Horizontal stripes. (c) Image (a)
degraded by horizontal stripes (b). (d) Vertical stripes. (e) Image (a) degraded by Vertical stripes (d). (f)–(j)
Singular values of (a)–(e).

3-5
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

and degraded OCTA images in figures 3.2(f) and (h) (or figure 3.2(j)) are similar and
their ranks are both equal to 256. We can conclude that the rank of OCTA image is
higher than the rank of stripes, and the low-rank constraint of stripe noise will not
affect the image content.
Total variation (TV) is a widely used regularizer for reducing noise while
preserving edges in images [33]. For OCTA images, anisotropic TV regularization
can obtain sharper boundaries in the clean image, which is important for subsequent
processing, such as vessel segmentation. As the stripe noise exists in the horizontal or
vertical direction, the anisotropic TV regularization can adapt these two situations
by loosening constraints on the stripe noise direction. The formulation of anisotropic
TV regularization is given as ∥C∥TV = ∥∇x C∥1 + ∥∇y C∥1, where ∥·∥1 is L1 norm
indicating the sum of the absolute value of the matrix elements. ∇x , ∇y are the row
and column derivative operators respectively.

3.2.1.3 Stripe noise removal model

As stated above, the low-rank constraint is used for the stripe component and the
anisotropic total variation (TV) regularization is used to achieve sharper boundaries
in the clean image, which will be important for segmenting vessels accurately. The
stripe noise removal model [32] is given as follows:
1 2
min C+ S − I F +τ C TV + λrank(S ) (3.2)
C,S 2
where τ and λ are two positive regularization parameters balancing the three terms.
Due to the non-convexity of the rank constraint, the nuclear norm is introduced to
replace the low-rank constraint as its convex substituted function [34]. The nuclear
norm based image decomposition model is given by
1 2
min C+S−I F +τ C TV +λ S *
(3.3)
C,S 2
The model (3.3) can optimize two variables C and S simultaneously by using an
alternatively minimizing strategy as follows.
(a) Stripe component S: Fixing other arguments, S is evaluated by minimizing
the following function:
1
Sˆ = argmin C+S−I 2
F +λ S *
(3.4)
S 2

which can be solved by the following soft-thresholding operation [35]:

⎧ k +1
(
⎪ S = U shrinkL* ∑,λ V
⎪
T
( ))
⎨ ⎧
⎪
⎛ ⎞⎫
⎪
(3.5)
⎪
( )
⎪shrinkL ∑,λ = diag⎨max⎜⎜∑−λ ,
* ⎪
⎝ ii
0⎟⎟⎬
⎪
⎠⎭i
⎩ ⎩

3-6
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

where Y − I = U ∑V T is the singular value decomposition of Y − I k , and

∑ii is the diagonal element of the singular value matrix ∑.
(b) Clean image C: Keeping other arguments fixed and minimizing with respect
to C , we have
1
Cˆ = argmin C+S−I 2
F +τ C TV (3.6)
C 2

where ∥C∥TV can be decomposed along with the directions x and y :

C TV = τx ∥∇x C∥1 + τy ∥∇y C∥1 (3.7)
where the weight τ along directions x and y is different due to the directional
characteristics of the stripe component. The alternative direction multiplier
method (ADMM) is introduced to solve the problem (3.6) efficiently. We
convert this problem into two sub-problems. Let Dx = ∇x C , Dy = ∇y C ,
D = [Dx , Dy ]T , ∇ = [∇x , ∇y ]T , τ = [τx, τy ]T , (3.6) equals the following
problem:
2
1 γ J
{Cˆ , Dˆ } = argmin C+S−I 2
F +τ D 1 + D − ∇C − (3.8)
C,D 2 2 γ F

The C -related sub-problem is followed by

2
1 γ J
Cˆ = argmin C+S−I 2
F + D − ∇C − (3.9)
C 2 2 γ F

We can obtain a closed form of (3.9) via the fast 2-D Fourier transform (FFT):
⎛  (I − S k+1 + ∇T (γ kD k − J k )) ⎞
C k +1 =  −1 ⎜ ⎟ (3.10)
⎝ 1 + γ k( (∇))2 ⎠

The D-related sub-problem is followed by

2
γ J
Dˆ = argmin τ D 1 + D − ∇C − (3.11)
D 2 γ F

which can be solved by a soft shrinkage operator

⎧ ⎛ Jk τ ⎞
⎪ D k+1 = shrinkL1⎜∇C k+1 + , ⎟
⎪ ⎝ γk γk ⎠
⎨ (3.12)
⎪shrinkL (r , ξ ) = r * max(r − ξ, 0).
⎪ 1
∣r ∣
⎩

3-7
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Then, we update the Lagrangian multipliers and penalization parameters as

follows:
J k+1 = J k + γ k(∇C k+1 − D k+1)
(3.13)
γ k +1 = γ k · ρ

3.2.2 The second stage: segmentation of vessels from stripe denoised images
Active contour models have demonstrated excellent performance in dealing with
challenging segmentation problems including vessel segmentation [17, 18]. They can
not only provide smooth and closed contours but also achieve subpixel accuracy on
vessel boundaries. The Global Minimization of the Active Contour/Snake model
(GMAC) [36] is introduced here for vessel segmentation. This model enhances the
well-known Chan–Vese (CV) model [15] and provides a convex solution for it.
GMAC obtains the segmentation results by detecting large image gradients and
homogeneous intensities regions. The GMAC model can be formulated as the
energy minimization problem below:

min
u, c1, c2
∫Ω g(x, y )∣∇u(x, y )∣dxdy + β∫Ω u(x, y )(C (x, y ) − c1)2dxdy
(3.14)
+β ∫Ω (1 − u(x , y ))(C (x , y ) − c2 )2 dxdy

where u(x , y ) is a characteristic function of a closed set ΩB , B represents the

boundaries of Ω , c1 and c2 are the average of C (x , y ) inside and outside of the
segmented region, respectively, and β is a small, positive balancing parameter. The
first term is the TV-norm of u(x , y ) with weighted edge indicator function g (x , y ). A
fast minimization method based on a dual formulation of the TV-norm [36] is
adopted to solve this problem.
(a) Region average intensity values c1, c2
Keeping other arguments fixed and minimizing with respect to c1 and c2 ,
respectively. We have the equations for c1 and c2 [15]:

∫Ω u(x, y )C (x, y )dxdy

c1 = (3.15)
∫Ω u(x, y )dxdy

∫Ω (1 − u(x, y ))C (x, y )dxdy

c2 = (3.16)
∫Ω (1 − u(x, y ))dxdy
which can be evaluated directly by giving the average intensities inside and
outside of the segmentation contour.

3-8
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

(b) Segmentation indicator function u

A convex regularization variational model is used according to [37]:

1 2
{uˆ , vˆ } = argmin u TV,g + u+v−C F
u, v 2θ
(3.17)
β β
+ (C − v)∥C − c1∥2F + (1 − C + v)∥C − c2∥2F
2 2

where the parameter θ is small so that we can approximate C = u + v. The function

u denotes the geometric information, while v represents the texture information in
the clean image C . Problem (3.17) can be solved by minimizing u and v iteratively.
The u -related sub-problem is followed by

1 2
uˆ = argmin u TV,g + u+v−C F (3.18)
u 2θ

The solution of (3.18) can be achieved efficiently by a fast dual projection algorithm
[38]. The derived solution is:

u k+1 = v k − θ divp k+1 (3.19)

where p is given by a fixed point method as follows:

p k + δt∇(div(p k ) − v / θ )
p k +1 =
δt (3.20)
1+ ∣∇(div(p k ) − v / θ )∣
g (x , y )

where δt ⩽ 1/8 is the temporal step.

The v-related sub-problem is followed by

1 2 β β
vˆ = u+v−C F + (C − v)∥C − c1∥2F + (1 − C + v)∥C − c2∥2F . (3.21)
2θ 2 2

The v-minimization can be achieved through the following update:

βθ
v k+1 = min{ max{u(x , y ) − [(C (x , y ) − c2 )2 − (C (x , y ) − c1)2 ], 0}, 1}. (3.22)
2
u in GMAC can be solved in a similar way.
Thus, the vessel segmentation from stripe removed images is achieved by solving
problem (3.3) to obtain the stripe removed and denoised image C , followed by
segmenting C by solving problem (3.14). The overall algorithm of the two stages
strategy is presented in algorithm 3.1.

3-9
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Algorithm 3.1. Segmenting images corrupted by stripe noises by the two stages
strategy (TSS).

Require: Image I with stripe noise, parameters τ , λ, β , γ , θ , ρ.

Ensure: Clean image C , stripe component S and segmentation indicator function u .
1: Initialize: set J 1 = 0, C1 = I , u1 = 0, v1 = 0
Stage 1: stripe noise removal
2: for k = 1: N do
3: Solve (3.5) for S k +1;
4: Compute C k +1 by solving (3.10);
5: Obtain D k +1 by solving (3.12);
6: Update Lagrangian multipliers J k +1 and penalization parameters γ k +1 via (3.13);
7: end for
Stage 2: segmentation of vessels from stripe denoised images
8: for k = 1: N do
9: Compute c1k +1 and c2 k +1 by solving (3.15) and (3.16);
10: Compute u k +1 by solving (3.19)
11: Compute v k +1 by solving (3.22)
12: end for

3.3 Joint destriping and segmentation of OCTA images (JDS)

In this section, a joint model [39] is given for simultaneously removing stripe
noise and segmenting images. The model tackles the problems of image
segmentation and considers the possible presence of stripe noise in a single
model. The joint model is formed by replacing the function C in the segmentation
model (3.4) with the stripe-removed and denoised image term. Furthermore, the
constraints of the decomposition problem (3.3) should be included in the joint
model. The illustration of the joint model is shown in figure 3.3. Building the
constraints into the segmentation model, the joint minimization model is
presented as follows:
1
min C + S − I 2F + τ C TV + λ S *
C , S , u, c1, c2 2
(3.23)
β β
+ α u TV,g + u ∥C − c1∥2F + (1 − u )∥C − c2∥2F
2 2
where ∥·∥TV,g is the TV-norm with weighted function, α , β are small, positive
balancing parameters. If α and β equal to 0, then model (3.23) converts to the
destriping model (3.3). The segmentation terms (fourth, fifth and sixth terms of
(3.23)) obtain the intensity and spatial information from the stripe removed image C .
To solve the joint model, each of the arguments is solved in turn. The detailed
optimization process is presented as follows.

3-10
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 3.3. Illustration of our joint model for segmentation images with stripe noise.

(a) Stripe component S

Keeping other arguments fixed, S is evaluated by minimizing the following
function:
1
Sˆ = argmin C+S−I 2
F +λ S *
(3.24)
S 2

(3.24) can be solved by the soft-thresholding operation [35]:

⎧ k +1
(
⎪ S = U shrinkL* ∑,λ V
⎪
T
( ))
⎨ ⎧
⎪
⎛ ⎞⎫
⎪
(3.25)
⎪
( ⎪
)
⎪shrinkL ∑,λ = diag⎨max⎜⎜∑−λ ,
*
⎝ ii
0⎟⎟⎬
⎪
⎠⎭i
⎩ ⎩

(b) Clean image C

Fixing other arguments and minimizing with respect to C , we have
1
Cˆ = argmin C + S − I 2F + τ C TV
C 2
(3.26)
β β
+ u ∥C − c1∥2F + (1 − u )∥C − c2∥2F
2 2

Problem (3.26) is converted into two sub-problems. Let Dx = ∇x C ,

Dy = ∇y C , D = [Dx , Dy ]T , ∇ = [∇x , ∇y ]T , τ = [τx, τy ]T , (3.26) equals the
following problem:

3-11
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

1
{C^ , D^} = argmin C+S−I 2
F +τ D TV
C,D 2
2 (3.27)
β 2 β 2 γ J
+ u C − c1 F + (1 − u ) C − c2 F + D − ∇C −
2 2 2 γ F

The C -related sub-problem is followed by

1 β
Cˆ = argmin C+S−I 2
F +τ C TV + u ∥C − c1∥2F
C 2 2
2 (3.28)
β γ J
+ (1 − u )∥C − c2∥2F + D − ∇C −
2 2 γ F

We can obtain a closed form of (3.28) via the fast 2-D Fourier transform (FFT):
⎛  (I − S k+1 + βu kc k + β(1 − u k )c k + ∇T (γ kD k − J k )) ⎞
C k+1 =  −1⎜ 1 2
⎟ (3.29)
⎝ 1 + β + γ k
(  ( ∇ )) 2
⎠

The D-related sub-problem is followed by

2
γ J
Dˆ = argmin τ D TV + D − ∇C − (3.30)
D 2 γ F

which can be solved by a soft shrinkage operator

⎧ ⎛ Jk τ ⎞
⎪ D k+1 = shrinkL1⎜∇C k+1 + , ⎟
⎪ ⎝ γk γk ⎠
⎨ (3.31)
⎪shrinkL (r , ξ ) = r * max(r − ξ, 0).
⎪ 1
∣r ∣
⎩

(c) Region average intensity values c1, c2

Fixing other arguments, c1 and c2 are evaluated as follows [15]:

∫Ω u(x, y )C (x, y )dxdy

c1 = (3.32)
∫Ω u(x, y )dxdy

∫Ω (1 − u(x, y ))C (x, y )dxdy

c1 = . (3.33)
∫Ω (1 − u ( x , y )) dxdy

3-12
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

(d) Segmentation indicator function u

With respect to u and fixing the other arguments, we have

β β
uˆ = argmin α u TV,g + u ∥C − c1∥2F + (1 − u )∥C − c2∥2F (3.34)
u 2 2

Similar to (3.17), a convex regularization variational model is adopted based

on [37]:
1 2
{uˆ , vˆ } = argmin α u TV,g + u+v−C F
u, v 2θ
(3.35)
β β
+ (C − v)∥C − c1∥2F + (1 − C + v)∥C − c2∥2F
2 2
where the parameter θ is small. Problem (3.35) can be solved by minimizing u and v
iteratively.
The u -related sub-problem is followed by
1 2
uˆ = argmin α u TV,g + u+v−C F (3.36)
u 2θ
The derived solution is given in the following:
u k+1 = v k − αθ divp k+1 (3.37)
where p is given by a fixed point method as follows:
p k + δt∇(div(p k ) − v /(αθ ))
p k +1 = .
δt (3.38)
1+ ∣∇(div(p k ) − v /(αθ ))∣
g (x , y )
The v-related sub-problem is followed by
1 2 β β
vˆ = u+v−C F + (C − v)∥C − c1∥2F + (1 − C + v)∥C − c2∥2F . (3.39)
2θ 2 2
The solution of (3.39) is given in the following:
βθ
v k+1 = min{ max{u(x , y ) − [(C (x , y ) − c2 )2 − (C (x , y ) − c1)2 ], 0}, 1}. (3.40)
2
Then, the Lagrangian multipliers and penalization parameters are updated as
follows:
J k+1 = J k + γ k(∇C k+1 − D k+1)
(3.41)
γ k +1 = γ k · ρ
The overall algorithm of joint model JDS is presented in algorithm 3.2.

3-13
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Algorithm 3.2. Segmenting images corrupted by stripe noises by the joint model
(JDS).

Require: Image I with stripe noise, parameters τ , λ, α, β, γ , θ , ρ.

Ensure: Clean image C , stripe component S and segmentation indicator function u .
1: Initialize: set J 1 = 0, C1 = I , u1 = 0, v1 = 0
2: for k = 1: N do
3: Solve S k +1 by using (3.25);
4: Compute C k +1 by solving (3.29);
5: Solve D k +1 using (3.31);
6: Obtain c1k +1 and c2 k +1 via (3.32) and (3.33);
7: Compute u k +1 by solving (3.37);
8: Solve (3.40) for v k +1;
9: Update Lagrangian multipliers J k +1 and γ k +1 via (3.41);
10: end for

3.4 Results
The proposed algorithms are evaluated in two parts: the effectiveness of destriping
and the effectiveness of segmentation. All experiments are run in MATLAB®
(R2018a) (Mathworks, MA) on a desktop with 8 GB RAM, Intel (R) Core (TM)
i5-7500 CPU @ 3.40 GHz.

3.4.1 Datasets
All the compared methods are extensively evaluated for their effectiveness and
efficiency for destriping and segmenting OCTA images. In this section, a brief
introduction of the used datasets is provided. All images used in this work were
collected with regulatory approvals and patients’ consent as appropriate.
(a) Real dataset: A real OCTA dataset consists of 85 images in the size of
840*840 that were collected from St Paul’s Eye Unit, Royal Liverpool
University Hospital, UK. Each image is taken in a 3 mm × 3 mm field of
view centered on the fovea from the internal limiting membrane (ILM) to
the inner plexiform layer (IPL), i.e., they are taken from superficial vascular
plexus (SVP) layer. An example is shown in figure 3.6(a). All images were
acquired using SPECTRALIS OCT (Heidelberg Engineering, Heidelberg,
Germany). The central lines of vessels in all images were manually
annotated using an in-house program by an ophthalmologist after proper
training on using the program, as shown in figure 3.4(b).
(b) Simulated dataset: All the OCTA images obtained in clinical settings are
corrupted by stripe noise. There is no clean image readily for the evaluation
of the performance. Following the general practice in image denoising and
enhancement, we created a synthetic dataset for comparison. A collection of
10 fluorescein angiography (FA) images from patients with diabetes by

3-14
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 3.4. Examples of synthetic and real datasets. The central lines in (b) are thickened for better
visualization.

HRA2 (Heidelberg Engineering, Heidelberg, Germany) was used as the

clean images. All the images have excellent quality as confirmed by visual
inspection of clinicians [40]. A region of 256 × 256 pixels centered at the
fovea was cropped and used in the experiments, as shown in figure 3.4(c).
This dataset is used to validate the effect of destriping.

3.4.2 Effectiveness of destriping

To test the effectiveness of destriping, two classical denoising methods (WMF and
WL), two destriping methods of other modalities (FFT and WFT) and two methods
described in this chapter (TSS and JDS) are compared and tested. These models are
briefly described as follows:
• WMF: The weighted median filter denoising method [41].
• WL: The wavelet denoising method [42].
• FFT: The fast Fourier transform based destriping method [27].
• WFT: The wavelet-Fourier combined transform based stripe artifact removal
method [26].
• TSS: The two-stage strategy presented in this chapter by destriping model
(3.3) followed by the segmentation model (3.14).
• JDS: The joint model (3.23) given in this chapter for destriping and
segmentation simultaneously.

Two evaluation metrics are utilized to test the above methods on simulated
dataset and real dataset, i.e., the peak signal-to-noise ratio (PSNR) and the
structural similarity (SSIM) [43]. The two metrics are calculated by using equations
(3.42) and (3.43), respectively.
MAXI 2
PSNR = 10 · log10( ) (3.42)
MSE

(2μx μy + a1)(2σxy + a2 )
SSIM (x , y ) = (3.43)
(μx + μy 2 + a1)(σx 2 + σy 2 + a2 )
2

3-15
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

where MAXI is the maximum pixel value of the noise-free image and MSE is the
mean squared error between the noise-free image and noisy image. x and y are two
windows of an image. μx , μy , σx 2, σy 2 are the average of x , y and the variance of x , y,
respectively. σxy is the covariance of x and y . a1, a2 are two variables.

3.4.2.1 Stripe direction influence

The stripe direction of OCTA image is usually either vertical or horizontal. The
effectiveness of stripe noise removal in two stripe directions is tested. We choose an
example simulated FA image and add the same stripes in the vertical and horizontal
directions separately, as shown in figures 3.5(b) and 3.6(b). Then six destriping
methods are used to remove the stripe noise, their visual results are presented in
figures 3.5(c)–(h) and figures 3.6(c)–(h). As WMF and WL are mainly designed for
random pepper noise or Gaussian noise, they fail to remove the stripe noise in both
vertical and horizontal directions. Although FFT and WFT can remove most of the
stripe noise, they generate some excessive destriping results with significant blurring
artifacts along the stripe direction. TSS and JDS can remove the stripes in both
directions and retain stripe-free areas well.
The corresponding quantitative results of all destriping methods are given in
tables 3.1 and 3.2 in terms of PSNR and SSIM, in which the highest PSNR and
SSIM values in each intensity scenario of simulated experiments are marked in bold.
Five different levels of stripes are added to the original image to test the robustness
of our methods. The intensity denotes the mean absolute value of the stripe lines.
TSS and JDS achieve better performance compared to other methods according to
tables 3.1 and 3.2. This demonstrates that the destriping model stated in this chapter
is effective to remove the stripe noise in both vertical and horizontal directions.

Figure 3.5. Simulated results of horizontal stripe noise removal. The intensity of the added stripe noise is 40 in
figure 3.5(b).

3-16
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 3.6. Simulated results of vertical stripe noise removal. The intensity of added stripe noise is 40 in
figure 3.6(b).

Table 3.1. Quantitative simulated results of horizontal stripe noise removal.

Stripe noise in the horizontal direction

Intensity = 10 Intensity = 20 Intensity = 30 Intensity = 40 Intensity = 50

Method PSNR SSIM PSNR SSIM PSNR SSIM PSNR SSIM PSNR SSIM

Degraded 26.65 0.636 21.24 0.360 18.29 0.230 16.33 0.161 14.90 0.120
WMF 26.78 0.624 21.27 0.377 18.15 0.254 16.09 0.187 14.60 0.147
WL 29.81 0.763 22.99 0.447 19.35 0.266 16.99 0.177 15.42 0.128
FFT 24.48 0.821 25.53 0.821 24.25 0.771 22.99 0.717 22.28 0.677
WFT 31.77 0.941 27.33 0.884 24.40 0.828 22.28 0.779 20.63 0.735
TSS 34.99 0.945 33.81 0.942 31.31 0.923 28.61 0.896 26.03 0.858
JDS 35.00 0.947 33.82 0.942 31.94 0.928 28.66 0.896 26.11 0.859

3.4.2.2 Simulated experiments

To simulate stripe noise in OCTA images, we will add some synthetic stripes to the
clean images in the simulated dataset. We only test horizontal stripes in this section
and the intensities of added stripes are the same as section 3.4.2.1. Table 3.3 reports
the destriping results in terms of PSNR and SSIM on the simulated dataset. Almost
all the methods can improve the PSNR and SSIM of degraded images after
destriping, while WMF and WL decrease the image quality when the added stripe
noise is weak (intensity equals 10). It can also be observed from table 3.3 that the

3-17
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 3.2. Quantitative simulated results of vertical stripe noise removal.

Stripe noise in the vertical direction

Intensity = 10 Intensity = 20 Intensity = 30 Intensity = 40 Intensity = 50

Method PSNR SSIM PSNR SSIM PSNR SSIM PSNR SSIM PSNR SSIM

Degraded 26.63 0.635 21.22 0.360 18.27 0.230 16.31 0.162 14.88 0.121
WMF 26.51 0.620 21.07 0.371 17.98 0.250 15.94 0.185 14.47 0.145
WL 29.65 0.744 23.11 0.425 19.36 0.249 16.96 0.164 15.29 0.118
FFT 22.85 0.793 23.67 0.796 23.60 0.765 23.09 0.724 22.98 0.697
WFT 32.76 0.947 28.55 0.898 25.23 0.840 22.91 0.788 21.14 0.743
TSS 34.69 0.955 33.45 0.946 31.77 0.930 28.87 0.901 26.36 0.862
JDS 34.70 0.955 33.57 0.949 31.78 0.931 29.04 0.902 26.40 0.862

Table 3.3. Quantitative results (PSNR (dB) and SSIM values) on the simulated dataset.

Stripe noise

Intensity = 10 Intensity = 20 Intensity = 30 Intensity = 40 Intensity = 50

Method PSNR SSIM PSNR SSIM PSNR SSIM PSNR SSIM PSNR SSIM

Degraded 29.36 0.848 23.74 0.634 20.72 0.480 18.73 0.376 17.29 0.304
WMF 26.59 0.701 23.93 0.592 21.51 0.500 19.61 0.422 18.10 0.366
WL 26.57 0.668 25.91 0.638 23.62 0.543 21.24 0.439 19.37 0.353
FFT 34.71 0.939 27.97 0.870 25.86 0.821 26.20 0.804 23.66 0.747
WFT 32.01 0.961 30.72 0.952 29.06 0.934 27.25 0.909 25.96 0.910
TSS 32.49 0.942 31.67 0.939 30.47 0.934 28.91 0.923 27.33 0.911
JDS 32.50 0.942 31.76 0.940 30.66 0.935 28.92 0.924 27.36 0.911

results of JDS outperform the competitors for all of the experiments under five
different levels of stripes.

3.4.2.3 Real experiments

The real OCTA stripe images are tested in this subsection. We choose a representa-
tive image (see figure 3.7) and zoom in four regions of this image, i.e. R1, R2, R3
and R4. It is shown that TSS and JDS can remove most stripe noise, while WMF
and WL can just remove some random noise, and FFT and WFT remove a little
stripe noise, but also add some noises in the image.

3.4.3 Effectiveness of segmentation

The effectiveness of segmentation is tested on two aspects: the destriping influence
on vessel segmentation and the vessel segmentation results. All the segmentation
models to be compared and tested in this section are introduced as follows:

3-18
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 3.7. Illustration of real OCTA destriping/denoising results. The first row shows the original OCTA
image and its four zoomed regions. The destriping results of WMF, WL, FFT, WFT, TSS and JDS are shown
from the second to seventh rows, respectively. The second to fifth columns show the zoomed regions R1, R2,
R3 and R4 of the whole image and their corresponding destriping results, respectively.

3-19
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

• CV: The Chan–Vese (CV) model of active contours without edges [15]
• GMAC: The Global Minimization of the Active Contour/Snake model
(GMAC) [36]
• TSS: The two-stage strategy presented in this chapter by the destriping model
(3.3) followed by the segmentation model (3.14).
• JDS: The joint model (3.23) given in this chapter for destriping and
segmentation simultaneously.

All of the vessel segmentation results are quantitatively evaluated on two metrics:
accuracy (Acc) and the area under the receiver operating characteristic curve (AUC)
[18]. In particular, AUC is a better metric for the overall performance measurement
for the imbalanced problem, i.e., in the vessel segmentation problem, the vessel
pixels are typically much fewer than the background pixels. The computations of
Acc and AUC are based on the centerline of the segmentation results and the
centerline annotation (see figure 3.4(b)). They are calculated as follows:
TP + TN
Acc = (3.44)
TP + FP + TN + FN

1 ⎛ TN TP ⎞
AUC = ⎜ + ⎟ (3.45)
2 ⎝ TN + FP TP + FN ⎠
where TP, FP, TN , FN indicate the true positive, false positive, true negative and
false negative, respectively.

3.4.3.1 Destriping influence on vessel segmentation

The destriping influence on vessel segmentation for the real OCTA dataset is
investigated in the following. This is an indirect method to show the effectiveness of
different destriping methods. Here we adopt two basic segmentation methods: CV
[7] and GMAC [14]. The source codes of these segmentation algorithms are acquired
from the authors or freely available on the Internet. These two segmentation
methods were utilized to segment the original noisy OCTA images and the
destriping images produced by different destriping methods (FFT, WFT, TSS and
JDS). WMF and WL are denoising methods and will not be tested in this section. It
should be noticed that we just utilize the destriping results for both TSS and JDS, the
segmentation results of these two methods will be compared in section 3.4.3.2. The
centerlines of the vessels from the segmentation were detected and the central lines of
the segmented SVP images are compared to the centerlines from manual annota-
tions. Quantitative results are given in table 3.4 in terms of Acc and AUC. After
destriping, almost all the Acc and AUC values improve, which demonstrates the
usefulness of destriping for vessel segmentation. JDS performs best in terms of Acc
and AUC. It improves Acc from 0.7851, 0.7705 to 0.7879, 0.7753 and AUC from
0.6367, 0.6221 to 0.6403, 0.6289, respectively. Visual comparison results are
illustrated in figure 3.8. It can be observed that the segmentation performance is
improved for all the segmentation methods after removing the stripe noise by JDS.

3-20
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 3.4. Destriping influence on vessel segmentation (CV) in terms of Acc and AUC.

CV GMAC

Method Acc AUC Acc AUC

Original 0.7851 0.6367 0.7705 0.6221

FFT 0.7825 0.6352 0.7706 0.6223
WFT 0.7819 0.6349 0.7705 0.6222
SD 0.7874 0.6389 0.7746 0.6287
JDS 0.7879 0.6403 0.7753 0.6289

Figure 3.8. Examples of the influence of destriping on vessel segmentation in real OCTA images by different
segmentation methods. From left to right: original image, a zoomed region of the original image, vessel
segmentation results of CV, a zoomed region of CV result, vessel segmentation results of GMAC, a zoomed
region of GMAC result. From top to bottom: before destriping, after destriping by JDS.

Table 3.5. Average accuracy values and CPU times of OCTA image segmentation.

Method Acc AUC Times

CV 0.7851 0.6367 177.82

GMAC 0.7705 0.6221 12.01
TSS 0.7743 0.6285 96.68
JDS 0.7877 0.6369 65.23

3.4.3.2 Vessel segmentation results

The average accuracy values and CPU times of TSS and JDS are listed in table 3.5.
The joint model JDS can achieve higher Acc and AUC compared to the other three
models, and spend less time than CV and TSS. We also give two segmentation
examples in figure 3.9. JDS can generate competitive segmentation results to TSS,
GMAC and CV efficiently.

3-21
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 3.9. Illustration of OCTA segmentation results.

3.5 Discussion and conclusions

OCTA is a new, non-invasive imaging technology that can provide blood vessel
maps at different depths. As an important element of retinal vasculature detection,
vessel segmentation separates a retinal vessel from its background and can be
applied in other vascular analysis. During the OCTA vessel segmentation, the
unavoidable stripe noise arose from the OCTA image acquisition affects the
accuracy of results. To address this issue, a two-stage strategy TSS and a joint
model JDS are presented in this chapter to segment retinal blood vessels in OCTA
images corrupted by stripe noise.
Both of TSS and JDS are based on the image decomposition and the active
contour theory. The TSS removed the stripe noise by a low-rank representation
model utilizing the low-rank constraints for stripes and TV regularization for
preserving vessel edges, and then segmented the retinal blood vessels by an active
contour model called GMAC. The JDS combined the models of stripe noise
removal and vessel segmentation. The performance of two models is tested on a
simulated FA dataset and a real OCTA dataset from the Royal Liverpool
University Hospital. The quantitative, visual and efficiency comparison results of
destriping and segmentation showed that both two models presented in this chapter
can solve the two problems efficiently, with JDS solving the problems simulta-
neously, faster and more accurately.
The stripe direction of the real OCTA image depends on its scan direction and is
usually either vertical or horizontal. TSS and JDS are able to remove stripe noise in
both two directions. The experimental results in section 3.4.2.1 showed their
effectiveness and practical value of both TSS and JDS with preferable PSNR and
SSIM values. To further prove the effectiveness of the destriping results, the results
of destriping influence on vessel segmentation are compared in section 3.4.3.1. The
results from four vessel segmentation methods proved that destriping is useful to
improve the segmentation performance and JDS can provide the highest improve-
ment compared to other methods in terms of Acc and AUC.

3-22
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

All parameters used in TSS and JDS are empirically chosen in the experiments.
The performance could be further improved if these parameters are fine ‘tuned’. On
the other hand, we have implemented the algorithms by using MATLAB®, which is
by no means optimal in terms of speed. Improvement can still be made by parallel
processing and GPU if needed.
In conclusion, this chapter presented a two-stage strategy TSS and a joint model
JDS for segmenting retinal blood vessels in OCTA images corrupted by stripe noise.
The experimental results demonstrate the effectiveness and efficiency of TSS and
JDS in comparison to classic denoising/destriping and segmentation models.
Besides, the destriping influence on vessel segmentation is also proved in this
chapter. It is believed that our work can inspire a way to consider the inclusion of
stripe removal in clinical vessel analysis models.

References
[1] Sheng B, Li P, Mo S, Li H, Hou X, Wu Q, Qin J, Fang R and Feng D D 2018 Retinal vessel
segmentation using minimum spanning superpixel tree detector IEEE Trans. Cybern. 99 1–13
[2] De-Carlo T E, Romano A, Waheed N K and Duker J S 2015 A review of optical coherence
tomography angiography (OCTA) Int. J. Retin. Vitr. 1 5
[3] Spaide R F, Fujimoto J G, Waheed N K, Sadda S R and Staurenghi G 2018 Optical
coherence tomography angiography Prog. Retin. Eye Res. 64 1–55
[4] Heimann H and Kellner U 2011 Atlas of Fundus Angiography (New York: Thieme)
[5] Venugopal J et al 2018 Repeatability of vessel density measurements of optical coherence
tomography angiography in normal and glaucoma eyes Br. J. Ophthalmol. 102 352–7
[6] Fingler J, Readhead C, Schwartz D and Fraser S E 2008 Phase-contrast OCT imaging of
transverse flows in the mouse retina and choroid Investig. Ophthalmol. Vis. Sci. 49 5055–9
[7] Jia Y, Tan O, Tokayer J, Potsaid B, Wang Y, Liu J J and Huang D 2012 Split-spectrum
amplitude-decorrelation angiography with optical coherence tomography Opt. Express 20
4710–25
[8] Tan A C S, Tan G S, Denniston A K, Keane P A, Ang M, Milea D, Chakravarthy U and
Cheung C M G 2018 An overview of the clinical applications of optical coherence
tomography angiography Eye 32 262–86
[9] Zhu C, Zou B, Zhao R, Cui J, Duan X, Chen Z and Liang Y 2017 Retinal vessel
segmentation in colour fundus images using extreme learning machine Comput. Med.
Imaging Graph. 55 68–77
[10] Martinez-Perez M E, Highes A, Stanton A V, Thorn S A, Chapman N, Bharath A A and
Parker K H 2002 Retinal vascular tree morphology: a semi-automatic quantication IEEE
Trans. Biomed. Eng. 49 912–7
[11] Niemeijer M, Van Ginneken B, Staal J, Suttorp-Schulten M S and Abramoff M D 2005
Automatic detection of red lesions in digital color fundus photographs IEEE Trans. Med.
Imaging 24 584–92
[12] Zana F and Klein J C 1999 A multimodal registration algorithm of eye fundus images using
vessels detection and Hough transform IEEE Trans. Med. Imaging 18 419–28
[13] Eladawi N, Elmogy M, Helmy O, Aboelfetouh A, Riad A, Sandhu H, Schaal S and El-Baz
A 2017 Automatic blood vessels segmentation based on different retinal maps from OCTA
scans. Comput. Biol. Med. 89 150–61

3-23
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[14] Eladawi N, Elmogy M, Khalifa F, Ghazal M, Ghazi N, Aboelfetouh A, Riad A, Sandhu H,

Schaal S and El-Baz A 2018 Early diabetic retinopathy diagnosis based on local retinal blood
vessel analysis in optical coherence tomography angiography (OCTA) images Med. Phys. 45
4582–99
[15] Chan T F and Vese L A 2001 Active contours without edges IEEE Trans. Image Process. 10
266–77
[16] Sum K and Cheung P Y 2008 Vessel extraction under non-uniform illumination: a level set
approach IEEE Trans. Biomed. Eng. 55 358–60
[17] Al-Diri B, Hunter A and Steel D 2009 An active contour model for segmenting and
measuring retinal vessels IEEE Trans. Med. Imaging 28 1488–97
[18] Zhao Y, Rada L, Chen K, Harding S P and Zheng Y 2015 Automated vessel segmentation
using infinite perimeter active contour model with hybrid region information with applica-
tion to retinal images IEEE Trans. Med. Imaging 34 1797–807
[19] Moult E et al 2014 Ultrahigh-speed swept-source OCT angiography in exudative AMD
Ophthalmic Surg. Lasers Imaging Retin. 45 496–505
[20] Zang P et al 2016 Automated motion correction using parallel-strip registration for wide-
field en face OCT angiogram Biomed. Opt. Express 7 2823–36
[21] Spaide R F, Fujimoto J G and Waheed N K 2015 Image artifacts in optical coherence
angiography Retina 35 2163
[22] Camino A, Zhang M, Gao S S, Hwang T S, Sharma U, Wilson D J, Huang D and Jia Y
2016 Evaluation of artifact reduction in optical coherence tomography angiography with
real-time tracking and motion correction technology Biomed. Opt. Express 7 3905–15
[23] Braaf B, Vienola K V, Sheehy C K K, Yang Q, Vermeer K A, Tiruveedhula P, Arathorn D
W, Roorda A and Boer J F D 2013 Real-time eye motion correction in phase-resolved OCT
angiography with tracking SLO Biomed. Opt. Express 4 51–65
[24] Zhang Q, Huang Y, Zhang T, Kubach S, An L, Laron M, Sharma U and Wang R K 2015
Wide-field imaging of retinal vasculature using optical coherence tomography-based micro-
angiography provided by motion tracking J. Biomed. Opt. 20 066008
[25] Kraus M F, Potsaid B, Mayer M A, Bock R, Baumann B, Liu J J, Hornegger J and
Fujimoto J G 2012 Motion correction in optical coherence tomography volumes on a per
A-scan basis using orthogonal scan patterns Biomed. Opt. Express 3 1182–99
[26] Münch B, Trtik P, Marone F and Stampanoni M 2009 Stripe and ring artifact removal with
combined wavelet-Fourier filtering Opt. Express 17 8567–91
[27] Shu-wen W C and Pellequer J L 2011 DeStripe: frequency-based algorithm for removing
stripe noises from AFM images BMC Struct. Biol. 11 7
[28] Liang X, Zang Y, Dong D, Zhang L, Fang M, Yang X, Arranz A, Ripoll J, Hui H and Tian
J 2016 Stripe artifact elimination based on nonsubsampled contourlet transform for light
sheet fluorescence microscopy J. Biomed. Opt. 21 106005
[29] Chang Y, Chen M, Yan L, Zhao X L, Li Y and Zhong S 2019 Toward universal stripe
removal via wavelet-based deep convolutional neural network IEEE Trans. Geosci. Remote
Sens. 58 2880–97
[30] Fehrenbach J, Weiss P and Lorenzo C 2012 Variational algorithms to remove stationary
noise: applications to microscopy imaging IEEE Trans. Image Process. 21 4420–30
[31] Sun L, Jeon B, Zheng Y and Wu Z 2017 Hyperspectral image restoration using low-rank
representation on spectral difference image IEEE Geosci. Remote Sens. Lett. 14 1151–55

3-24
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[32] Chang Y, Yan L, Wu T and Zhong S 2016 Remote sensing image stripe noise removal: From
image decomposition perspective IEEE Trans. Geosci. Remote Sens. 54 7018–31
[33] Kamilov U S 2016 A parallel proximal algorithm for anisotropic total variation minimiza-
tion IEEE Trans. Image Process. 26 539–48
[34] Fazel M 2002 Matrix rank minimization with applications PhD Thesis Stanford University,
Stanford, CA
[35] Cai J F, Candes E J and Shen Z 2010 A singular value thresholding algorithm for matrix
completion SIAM J. Optim. 20 1956–82
[36] Bresson X, Esedolu S, Vandergheynst P, Thiran J P and Osher S 2007 Fast global
minimization of the active contour/snake model J. Math. Imaging Vis. 28 151–67
[37] Aujol J F, Gilboa G, Chan T and Osher S 2006 Structure-texture image decomposition
modeling, algorithms, and parameter selection Int. J. Comput. Vision 67 111–36
[38] Chambolle A 2004 An algorithm for total variation minimization and applications J. Math.
Imaging Vis. 20(1-2) 89–97
[39] Wu X, Gao D, Williams B M, Stylianides A and Zheng Y 2019 Joint destriping and
segmentation of OCTA images Annual Conf. on Medical Image Understanding and Analysis
pp 423–35
[40] Zheng Y, Gandhi J S, Stangos A N, Campa C, Broadbent D and Harding S P 2010
Automated segmentation of foveal avascular zone in fundus fluorescein angiography
Investig. Ophthalmol. Vis. Sci. 51 3653–9
[41] Brownrigg D 1984 The weighted median filter Commun. ACM 27 807–18
[42] Chang S G, Yu B and Vetterli M 2000 Adaptive wavelet thresholding for image denoising
and compression IEEE Trans. Image Process. 9 1532–46
[43] Wang Z, Bovik A C, Sheikh H R and Simoncelli E P 2004 Image quality assessment: from
error visibility to structural similarity IEEE Trans. Image Process. 13 600–12

3-25
 IOP Publishing

Photo Acoustic and Optical Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
Ayman El-Baz and Jasjit S Suri

Chapter 4
Optical coherence tomography angiography
changes in early type 3 neovascularization after
anti-vascular endothelial growth factor
treatment
Riccardo Sacconi, Carlo Di Biase, Enrico Borrelli, Lea Querques, Francesco
Prascina, Ilaria Zucchiatt, Francesco Bandello and Giuseppe Querques

Type 3 macular neovascularization (MNV) is a peculiar form of neovascular age-

related macular degeneration (AMD) characterized by a predilection for the neuro-
sensory retina. In the early stages, type 3 MNV is characterized by the presence of
neovascular capillaries detected mainly in the deep vascular complex (DVC) and the
avascular slabs, without connection with the retinal pigment epithelium (RPE). At this
stage, type 3 MNV is characterized by the absence of intraretinal fluid. Over time,
typically, nascent type 3 MNV displays a down-growth progression from the DVC to
the RPE. Only when the neovascular complex reaches the RPE and sub-RPE space
does the type 3 neovascularization become ‘active’ and it is complicated by intraretinal
fluid. Due to the ‘aggressive’ nature of type 3 neovascularization, it is very important
to treat this kind of neovessel in the early stages of this lesion, offering thus hope for
more favorable results. The treatment with anti-vascular endothelial growth factor
(VEGF) injections should be started when the complete progression of nascent type 3
MNV (i.e. from the DVC to the RPE) is detected.
After anti-VEGF treatment, the detectable flow of type 3 MNV is no longer
continuous with the RPE and sub-RPE space, and thus remains isolated in the
retinal layers, featuring absence typical exudative signs of type 3 MNV. However, at
the time of recurrence, type 3 MNVs show the presence of intra/sub-retinal
exudation with the restoration of the flow deepening from the DVC to the RPE/
sub-RPE space. This restoration is therefore featuring the early phases of recurrent
exudation in type 3 MNVs. Thus, the communication of the neovascular complex

doi:10.1088/978-0-7503-2060-3ch4 4-1 ª IOP Publishing Ltd 2021

 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

with the RPE and sub-RPE space from the DVC seems necessary to develop an
‘active’ form of type 3 MNV in both nascent type 3 MNV and recurrent type 3
MNV after treatment.

4.1 Introduction
Type 3 macular neovascularization (MNV) is a particular kind of neovascular age-
related macular degeneration (AMD) with neurosensory retina [1] predilection. It
has been initially described by Hanett et al [2] as an ‘Abnormal deep retinal vascular
complex’ in patients with AMD. Subsequently, several study groups have speculated
about the genesis of this type of MNV. A choroidal origin has been suggested by
Gass et al [3] who defined these MNVs as ‘occult chorio-retinal anastomosis’.
Contrarily, Yannuzzi et al [4] hypothesized an intraretinal origin, proposing the term
‘retinal angiomatous proliferation’ (RAP). On the other hand, Freund et al [1]
coined the term ‘type 3 MNV’ referring to any sign of intraretinal MNV in AMD,
without considering its retinal or choroidal origin. The term has been proposed so as
to expand the Gass classification of MNV on the basis of the anatomical site. The
evolution of type 3 MNV has been studied by different groups with the use of high-
resolution imaging techniques. Their evaluations have supported the intraretinal
origin of this common form of MNV in patients with AMD [5–11]. Since no
histopathological evidence of retinal neovessels and choriocapillaris anastomosis has
been found, novel imaging techniques suggested a crucial role of the deep vascular
complex (DVC) in the pathogenesis of the disease. Optical coherence tomography
angiography (OCT-A), has been of central importance defining the structure of type
3 MNV. In this latter, OCT-A shows an intraretinal vascular complex deriving from
the DVC which could be frequently associated with abutting telangiectatic vessels
[12–14]. Early identification of type 3 is of primary importance in order to perform
an immediate therapy and limit late stage disease progression.

4.2 Nascent type 3 MNV

Nascent type 3 macular neovascularization has been recently identified by our study
group. It is characterized by hyperreflective foci in the retinal layers (HRF) at
structural OCT with a noticeable flow at OCT-A [15]. Therefore, OCT-A is capable
of neovascular flow detection earlier than the clinical outbreak of type 3 MNV and
prior to the observation of exudation (e.g. intraretinal fluid at structural OCT) [15].
In our series, the presence of neovascular flow within the HRF was observed at
structural OCT in all cases. Moreover, every case evolved to an exudative form of
type 3 MNV. In a past publication, Su et al [16] explained the appearance of HRF
using structural OCT as a first sign of type 3 MNV ascribing intraretinal foci to
migration of retinal pigment epithelial (RPE) cells. In another study, the presence of
HRF at spectral domain OCT corresponds to a small capillary bunch if evaluated
using OCT-A, although only posterior to the detection of intraretinal edema [17–19].
Even if HRF could correspond to migrated RPE cells, our group disclosed the
existence of detectable flow inside these foci even in the lack of exudate, thus an
‘active’ type 3 MNV forms. The landmark of the nascent type 3 MNV is HRF

4-2
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

downgrowth advancement, with flow starting from the DVC toward the RPE. For
this reason, we described that type 3 MNVs become ‘active’ (i.e. presence of
intraretinal fluid) only at the time of detectable flow between the DVC to the RPE/
sub-RPE space.
In the early phases of nascent type 3, neovascularizations are seen as HRF with
detectable flow noticeable exclusively in the DVC and the avascular slabs, and not
consecutive to the RPE. At this stage, nascent type 3 MNV displays the lack of
intraretinal fluid. Over time, commonly, nascent type 3 MNV shows a downgrowth
advancement from the DVC to the RPE. Type 3 MNV becomes ‘active’ and it is
characterized by intraretinal fluid, solely when the MNV reaches the RPE and sub-
RPE space. OCT-A is of predominant significance, in this process, to recognize the
early stages of the disease, changing our treatment and follow-up approach.
The retinal derivation theory of type 3 MNV is supported by the OCT-A findings
of nascent type 3 neovascularization. In fact, prior to leakage, nascent type 3
undeniably originates from the DVC over a drusenoid pigment epithelial detach-
ment (PED) and subsequently advances into the RPE and sub-RPE space generating
intraretinal exudation.
Still, more rarely, the intraretinal neovascularization has been related to simulta-
neous type 1 MNV below the drusenoid PED, and no evident anastomosis amid
these findings.
Miere et al [28] interestingly theorized that the existence of a drusenoid PED
related to the loss of photoreceptors and to the atrophy of the outer retinal layers
and of the RPE may stimulate the formation of a DVC neovascularization with
latter down-growth towards the RPE/sub-RPE space.
Our results likewise support this hypothesis.
In our series, several HRFs with detectable flow using OCT-A did not evolve to
‘active’ type 3 MNV during follow-up, displaying only an isolated flow to the deep
capillary plexus (DCP) and avascular slab.
Furthermore, we also published an HRF case showing signs of flow that
displayed a disappearance of the lesion during the follow-up, using fluorescein
angiography, indocyanine green angiography and OCT-A.
For this matter, it is noteworthy to stress that not every HRF with flow advance
to an exudative form of type 3 MNV, but exclusively HRFs with flow that display a
progressive growth starting from the deep capillary plexus and reaching the RPE/
sub-RPE space.
It is strategic to treat these types of neovessels in the initial phases of this lesion,
owing to the ‘aggressive’ nature of type 3 MNV, hence offering hope for more
favorable results [22, 28].
Even so, we need to know the correct timing of when to start treatment.
We should underline that timely diagnosis of HFR with flow owing to migrating
of the RPE cells inside the retinal layers is consequentially more frequently met than
HRF by nascent type 3 MNV. The cardinal dissimilarity among these two types is
that, in the case of migrating RPE cells, hyper-reflective foci may seldom show
down-growth advancement.

4-3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

A higher possibility of downward advancement was suggested with OCT-A signal

of flow matching with HRF, while nascent type 3 neovascularization infrequently
naturally reverted without treatment unless there was progress into the RPE/sub-
RPE space.
Owing to this possibility of a nascent form of type 3 MNV regression, our
suggestion is no treatment until the neovascularization progresses as far as the RPE/
sub-RPE space.
The diagnosis of nascent type 3 MNV should accordingly authorize an earlier
follow-up using OCT-A in order to record the downgrowth towards the RPE/sub-
RPE space.
Our suggestion is that nascent form of type 3 MNV complete progression (i.e.
starting from the DVC reaching the RPE) may point to the necessity of prompt
treatment.

4.3 Treatment of active type 3 MNV

In various MNV types, multiple clinical studies have evaluated the outcomes of anti-
VEGF treatment [20–25].
Ebenezer et al engaged in a two-year clinical outcome trial, ‘Comparison of
AMD Treatments Trials (CATT)’ [22]. The study assessed 126 eyes with subtype 3
MNV. Daniel et al disclosed results that type 3 MNV eyes treated using anti-VEGF
injections required fewer injections in relationship to the type 1 and 2 MNV subtypes
(6.1 versus 7.4 during the first year and 5.4 versus 6.6 during the second year).
Moreover, at one-year follow-up, the average BCVA improvement was better in
type 3 NV patients than in other types of MNV (10.6 versus 6.9 letters) [22].
Even so, at follow-up finish, type 3 lesions displayed exhibited an augmented
incidence of RPE atrophy [22].
In a real-life practice series completed in seven Italian centers, Parodi et al [23]
confirmed that fewer anti-VEGF treatments (bevacizumab or ranibizumab) are
needed in type 3 MNV than in other MNV subtypes.
Specifically, 95 patients had undergone an average of 4.4 injections during the
first year of treatment recording a considerable upgrading of the BCVA (from 0.66
to 0.53 logMAR). [23]
Different papers have presented the value of ranibizumab or bevacizumab in the
treatment of type 3 NV, while less is specified about the efficacy of aflibercept.
Cho et al [26] studied the efficiency of aflibercept in relationship to ranibizumab
analyzing 63 type 3 MNV naïve eyes treatments. Their conclusions showed similar
BCVA improvement in the two groups at one year follow-up.
Nevertheless, they showed an increased RPE atrophy occurrence in the afliber-
cept group [26].
In fact, Matsumoto et al [27] presented comparable improvements in BCVA in
type 3 MNV treated with a treat-and-extend protocol based on aflibercept.
Mean BCVA by ETDRS chart was 9.5 letters, while central macular thickness fell
from 340 μm to 133 μm. In a recently published paper [28], the authors observed

4-4
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

improvements after three sequential aflibercept injections in 47% and vision stability
in 42% of the examined eyes.
The therapy was less effective presumably owing to a higher incidence of type 3
lesions in the advanced forms (73% of the examined eyes showed PED).
Summarizing, both aflibercept and ranibizumab injections, in terms of BCVA
and central macular thickness improvement, obtained good results in type 3 NV
treatment.
The treatment goal is early stage disease patient management to reduce treatment
during follow-up obtaining a better VA.
Nevertheless, to minimize atrophy development, we must be cautious with the
number of injections.

4.4 OCT-A features of recurrent type 3 MNV after anti-VEGF

injections
In a current study, our group examined OCT-A early changes characterizing post
treatment recurrent type 3 MNV.
This is especially practical to finding implicit imaging bio-markers for an
immediate re-treatment that may likely preclude advancement to advanced stages
of the disease.
After visiting 12 post anti-VEGF treated patients, we saw that at the post
treatment non-exudative stage, all lesions incompletely reversed, in fact,
detectable flow was observed exclusively in the Deep Vascular Complex and/or
avascular slab (no detectable flow at or beneath the RPE).
Hence, in all patients, the appreciable flow was then found disjointed with the
RPE/sub-RPE space, and thus lasted apart within the retinal layers, underlining lack
of type 3 MNV characteristic exudative signs (figures 4.1 and 4.2).
While, during relapse, all type 3 MNVs displayed the appearance of intraretinal
and or subretinal exudation with the regeneration of the flow out of the DVC
towards the RPE space. This regeneration, present in each case, was thus an early
phase characteristic of type 3 MNV recurrent exudation (figures 4.1 and 4.2).
Therefore, the neovascular complex relationship between the deep vascular
complex and the RPE/sub-RPE space appears essential to evolve to an ‘active’
pattern of type 3 MNV in both nascent type 3 MNV and post treatment recurrent
type 3 MNV.
Outcomes about naïve eyes treatment by anti-VEGF shots in type 3 MNVs
secondary to AMD were previously published by Miere et al [29].
Two different OCT-A scenes were described. The initial one presents eyes with
exudation at one-year examination showing duration of a well-defined lesion in the
external retinal layers and choriocapillaris.
However, the latter one is characterized by eyes showing a complete exudate
resolution at 12-month follow-up with a total loss of the early tuft-shaped and claw-
like flow signals, without outer retina and choriocapillaris level flow, regardless of
the DVC feeder vessels permanence.

4-5
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 4.1. Multimodal imaging evaluation including (a) multicolor image, (b) fundus autofluorescence, and
(c) optical coherence tomography angiography (OCT-A) of the left eye of a patient affected by type 3 macular
neovascularization. In detail, enface OCT-A of the (c) avascular slab and (D) RPE-RPEfit at the baseline
showed the neovascular network. Cross-sectional B-scan with flow showed the presence of a lesion growing
from the DVC to the RPE space at the baseline, with a regression after the anti-VEGF treatment, and a
restoration at the time of the recurrence.

Our study results were in agreement with Miere et al. In our study, naïve
treatment of type 3 neovascularization displayed, at baseline, neovessels rising from
the DVC to the RPE/sub-RPE. This OCT-A detected linkage was lost in all cases
post intra-subretinal exudation resolution with the neovascular tuft persistency at
the DVC and/or avascular slab level. However, regeneration of the flow starting
from the deep vascular plexus to the RPE/sub-RPE defined the recurrence of type 3
MNV related exudative signs.
In a previous histological study, Skalet et al [25] presented the eye globe findings
of a 93 year-old male patient with type 3 MNV treated by several anti-VEGF shots.
They reported that ranibizumab had been unsuccessful in reducing the type 3 lesion
entirely.
Actually, regardless of no exudative signs, the type 3 lesion showed up as thick-
walled neovessels in the retinal layers. Remarkably at the RPE level, with no signs of
neovascular tissue, a thin band of hyperreflective tissue underlying the type 3 lesion
was found [25]. The previous histopathologic findings were confirmed by our in vivo
OCT-A findings.
Our group reported that type 3 neovascularization did not completely dissolve
after anti-VEGF injections. Lesion regression was obtained with the vanishing of
recognizable flow at the outer retina level and without evidence of a link with the
RPE, even if a neovascular flow was still visible at the DVC. At recrudescence, an
OCT-A detectable flow remerged at the outer retina level with a neovascularization

4-6
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 4.2. Optical coherence tomography angiography (OCT-A) evaluation of the right eye of a patient with
a type 3 macular neovascularization. En face OCT-A of the avascular slab (first row) and of the RPE-RPEfit
slab (second row), and cross-sectional B-scan with flow (third row) showed the neovascular network with the
presence of flow growing from the DVC to the RPE space at the baseline, with a regression after the anti-
VEGF treatment, and a restoration at the time of the recurrence.

down-growth advancement from the DVP to the RPE. In concert with former
histopathological records and up to date in vivo OCT-A results, we embrace the
thesis that type 3 lesions do not dissolve after anti-VEGF shots, since the lesion
between the DVC and the RPE/sub-RPE space is still present as a fibrous tissue
under the DVC neovascular complex. OCT-A image lesion disappearance at the
RPE and sub-RPE space level may be associated both with the lack of or with a very
sluggish flow (i.e. no flow observable at OCT-A) within the hyperreflective tissue of
the type 3 neovascularization post anti-VEGF injections. Nonetheless, at recrudes-
cence, there might be an incrementing flow inside the type 3 MNV that defines the
recovered flow in the entire neovascularization from the DVC to the RPE.
Post treatment, type 3 MNVs show no exudative signs before the retinal
neovascular complex flow does not advance de novo from the DVC into the RPE.
When this happens, type 3 MNV recrudescence is characterized by exudation (i.e.
intraretinal fluid), prefiguring the ‘active’ form of type 3 neovascularization. This
linkage amidst the DVC and the RPE appears to be necessary for new exudation
production secondary to recrudescence type 3 MNV. Consequently, the discovery of

4-7
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

appreciable flow down-growth advancement from the DVC to the RPE, indicates
impending recrudescence and may encourage prompt retreatment. Amidst these
lines, OCT-A could be intended as a new useful imaging technique to determine
initial signs of recrudescence, in reality prior to exudative signs found by structural
OCT.

References
[1] Freund K B, Ho I V and Barbazetto I A et al 2008 Type 3 neovascularization: the expanded
spectrum of retinal angiomatous proliferation Retina 28 201–11
[2] Hartnett M E, Weiter J J, Garsd A and Jalkh A E 1992 Classification of retinal pigment
epithelial detachments associated with drusen Graefes. Arch. Clin. Exp. Ophthalmol 230 11–9
[3] Gass J D, Agarwal A, Lavina A M and Tawansy K A 2003 Focal inner retinal hemorrhages
in patients with drusen: an early sign of occult choroidal neovascularization and chorior-
etinal anastomosis Retina 23 741–51
[4] Yannuzzi L A, Negrão S and Iida T et al 2001 Retinal angiomatous proliferation in age-
related macular degeneration Retina 21 416–34
[5] Miere A, Querques G, Semoun O, El Ameen A, Capuano V and Souied E H 2015 Optical
coherence tomography angiography in early type 3 neovascularization Retina 35 2236–41
[6] Miere A, Querques G and Semoun O et al 2017 Optical coherence tomography angiography
changes in early type 3 neovascularization after anti-vascular endothelial growth factor
treatment Retina 37 1873–9
[7] Querques G, Souied E H and Freund K B 2013 Multimodal imaging of early stage 1 type 3
neovascularization with simultaneous eye-tracked spectral-domain optical coherence tomog-
raphy and high-speed real-time angiography Retina 33 1881–7
[8] Querques G, Souied E H and Freund K B 2015 How has high-resolution multimodal
imaging refined our understanding of the vasogenic process in type 3 neovascularization?
Retina 35 603–13
[9] Su D, Lin S and Phasukkijwatana N et al 2016 An updated staging system of type 3
neovascularization using spectral domain optical coherence tomography Retina 36 S40–9
[10] Tan A C, Dansingani K K, Yannuzzi L A, Sarraf D and Freund K B 2017 Type 3
neovascularization imaged with cross-sectional and en face optical coherence tomography
angiography Retina 37 234–46
[11] Freund K B, Zweifel S A and Engelbert M 2010 Do we need a new classification for
choroidal neovascularization in age-related macular degeneration? Retina 30 1333–49
[12] Miere A, Querques G and Semoun O et al 2015 Optical coherence tomography angiography
in early type 3 neovascularization Retina 35 2236–41
[13] Kuehlewein L, Dansingani K K and de Carlo T E et al 2015 Optical coherence tomography
angiography of type 3 neovascularization secondary to age-related macular degeneration
Retina 35 2229–35
[14] Miere A, Querques G and Semoun O et al 2017 Optical coherence tomography angiography
changes in early type 3 neovascularization after anti-vascular endothelial growth factor
treatment Retina 37 1873–9
[15] Sacconi R, Sarraf D and Garrity S et al 2018 Nascent type 3 neovascularization in age-
related macular degeneration Ophthalmol. Retina 2 1097–106
[16] Su D, Lin S and Phasukkijwatana N et al 2016 An updated staging system of type 3
neovascularization using spectral domain optical coherence tomography Retina 36 S40–9

4-8
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[17] Kuehlewein L, Dansingani K K and de Carlo T E et al 2015 Optical coherence tomography

angiography of type 3 neovascularization secondary to age-related macular degeneration
Retina 35 2229–35
[18] Miere A, Querques G, Semoun O, El Ameen A, Capuano V and Souied E H 2015 Optical
coherence tomography angiography in early type 3 neovascularization Retina 35 2236–41
[19] Dansingani K K, Naysan J and Freund K B 2015 En face OCT angiography demonstrates
flow in early type 3 neovascularization (retinal angiomatous proliferation) Eye 29 703–6
[20] Park Y G and Roh Y J 2015 One year results of intravitreal ranibizumab monotherapy for
retinal angiomatous proliferation: a comparative analysis based on disease stages BMC
Ophthalmol 15 182
[21] Tsaousis K T, Konidaris V E and Banerjee S et al 2015 Intravitreal aflibercept treatment of
retinal angiomatous proliferation: a pilot study and short-term efficacy Graefes. Arch. Clin.
Exp. Ophthalmol 253 663–5
[22] Daniel E, Shaffer J and Ying G S et al 2016 Comparison of age-related macular degeneration
treatments trials (CATT) research group. Outcomes in eyes with retinal angiomatous
proliferation in the comparison of age-related macular degeneration treatments trials
(CATT) Ophthalmology 123 609–16
[23] Parodi M B, Donati S and Semeraro F et al 2017 Intravitreal anti-vascular endothelial
growth factor drugs for retinal angiomatous proliferation in real-life practice J. Ocul.
Pharmacol. Ther 33 123–7
[24] Inoue M, Arakawa A, Yamane S and Kadonosono K 2014 Long-term results of intravitreal
ranibizumab for the treatment of retinal angiomatous proliferation and utility of an
advanced RPE analysis performed using spectral-domain optical coherence tomography
Br. J. Ophthalmol 98 956–60
[25] Skalet A H, Miller A K, Klein M L, Lauer A K and Wilson D J 2017 Clinicopathologic
correlation of retinal angiomatous proliferation treated with ranibizumab Retina 37 1620–4
[26] Cho H J, Hwang H J, Kim H S, Han J I, Lee D W and Kim J W 2018 Intravitreal aflibercept
and ranibizumab injections for type 3 neovascularization Retina 38 2150–8
[27] Matsumoto H, Sato T and Morimoto M et al 2016 Treat-and-extend regimen with
aflibercept for retinal angiomatous proliferation Retina 36 2282–9
[28] Chou H D, Wu W C, Wang N K, Chuang L H, Chen K J and Lai C C 2017 Short-term
efficacy of intravitreal aflibercept injections for retinal angiomatous proliferation BMC
Ophthalmol 17 104 Erratum in: BMC Ophthalmol 2017;17 (1):144
[29] Miere A, Querques G and Semoun O et al 2017 Optical coherence tomography angiography
changes in early type 3 neovascularization after anti-vascular endothelial growth factor
treatment Retina 37 1873–9

4-9
 IOP Publishing

Photo Acoustic and Optical Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
Ayman El-Baz and Jasjit S Suri

Chapter 5
Optical coherence tomography angiography in
multiple sclerosis and neuromyelitis optica
Olwen C Murphy and Shiv Saidha

The anterior visual pathway is a key site of disease in both multiple sclerosis (MS)
and neuromyelitis optica spectrum disorder (NMOSD). The retina is considered a
‘window into the brain’ and optical coherence tomography (OCT) is now a widely
used imaging tool for evaluation of the anterior visual pathway in MS and
NMOSD. Thinning of the peripapillary retinal nerve fiber layer (pRNFL) and
ganglion cell + inner plexiform layer (GCIPL) reflecting neuroaxonal loss are key
retinal changes known to occur in MS and NMOSD as a result of acute optic
neuritis (ON), subclinical optic neuropathy, or a combination of both. In MS, retinal
layer thinning is considered an imaging biomarker of neurodegeneration that can be
assessed in a rapid, reproducible manner using OCT. In recent years the advent of
OCT angiography (OCTA) has offered the opportunity to explore metabolic
demand and vascular integrity in the retina in these diseases using a non-invasive
depth-resolved technique. Studies using OCTA have demonstrated that retinal
vascular plexus densities are reduced in people with MS (in ON eyes more than
non-ON eyes), and that greater reductions in vascular plexus densities correlate with
longer MS disease duration, higher levels of disability and poorer visual acuity.
Reductions in retinal vascular plexus densities have also been identified with OCTA
in NMOSD—particularly in ON eyes. We review and interpret the growing body of
research utilizing OCTA in MS and NMOSD, and discuss some limitations of
OCTA imaging in the context of these diseases.

5.1 Introduction
The anterior visual pathway is an important site of disease in demyelinating
disorders of the central nervous system (CNS). Around half of patients with multiple
sclerosis (MS) will experience a clinical episode of acute optic neuritis (ON) at some

doi:10.1088/978-0-7503-2060-3ch5 5-1 ª IOP Publishing Ltd 2021

 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

point in their disease course, and subclinical optic neuropathy is almost ubiquitous,
with demyelinated plaques identified in the optic nerves of up to 99% of people
with MS at post-mortem [1, 2]. Neuromyelitis optica spectrum disorder (NMOSD)
also frequently affects the anterior visual pathway, with ON being the inaugural
manifestation of disease in around 40% of aquaporin-4 antibody seropositive
patients [3, 4]. ON and subclinical optic neuropathy cause axonal injury and
degeneration within the optic nerve, which may proceed anterogradely (towards
the thalamus and posterior visual pathway), and retrogradely (towards the
retina). Ultimately, retrograde neuroaxonal degeneration can result in thinning
of the retinal nerve fiber layer (RNFL), and loss of retinal ganglion cell bodies—
which is seen histopathologically in approximately 80% of MS eyes at post-
mortem [5, 6]. Research using optical coherence tomography (OCT) has offered a
useful tool for measuring these dynamic retinal processes in vivo.
Retinal tissue is highly metabolically active and highly vascularized (with a
distinctly-structured vascular network), and represents a useful accessible location
for examining these processes. Neuroaxonal degeneration in the anterior visual
pathway in MS and NMOSD is likely to result in reduced metabolic demand in
retinal tissue, and exploring retinal vascular metrics can offer useful information
regarding metabolic demand and vascular supply. Moreover, aberrant vascular
processes are known to be extensive in patients with MS, including excess platelet
activation, increased thrombophilic markers, altered blood–brain-barrier (BBB)
permeability, endothelial cell dysfunction, hypoxia-like tissue injury, perivascular
iron deposition, and vascular occlusion within lesions [7–12]. While pathobiologic
processes differ in NMOSD as compared to MS, vascular processes are also thought
to be at play in the NMOSD disease state; altered BBB permeability is a
pathological hallmark of the disease, and NMOSD lesions are characterized by
perivascular immunoglobulin deposition, perivascular complement activation, and
blood vessel thickening [13]. Brain imaging studies have suggested that reductions in
cerebral perfusion are present in patients with MS and NMOSD [14–18], and that
MS lesions have a predilection for peri-venular locations [19, 20]. Many of these
pathobiological processes occurring in the brain may be capitulated in the retina.
Indeed, post-mortem studies of MS eyes have shown that venular thickening, peri-
vascular inflammation, peri-venular gliosis, and alterations in endothelial tight
junctions are present in the retina and optic nerve head [5, 21]. While similar
histopathological studies of the retina are lacking in NMOSD, aquaporin-4 (the
target of the pathogenic antibody identified in most patients with NMOSD) is highly
expressed in the retina by astrocytes and Müller glial cells [22], and retinitis has been
demonstrated in animal models of NMOSD [23]. In vivo imaging studies using
fluorescein angiography or fundus photography have suggested the presence of
retinal peri-phlebitis in some MS eyes [24–26], and the presence of retinal focal
arteriolar narrowing and attenuation of the peripapillary vascular tree in NMOSD
eyes [27].
The advent of OCT angiography (OCTA) in recent years has offered a new
technique to explore the retinal vasculature in vivo. The combination of OCT and
OCTA may allow the concomitant evaluation of both neurodegenerative and

5-2
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

vascular processes operative in MS and NMOSD. Furthermore, insights gained

from the retina may reflect global disease processes. For these reasons, there has
been substantial interest in employing OCTA for research in MS and NMOSD. In
this chapter, we will focus on the background provided by extensive studies of OCT-
derived retinal layer thicknesses in MS and NMOSD, the expanding body of
research employing OCTA in these disorders, and the potential limitations of OCTA
as an imaging technique.

5.2 Optical coherence tomography (OCT) of the retina

Retinal layer thicknesses can be measured in vivo using optical coherence tomog-
raphy (OCT)—a rapid, non-invasive, well-tolerated, inexpensive, reproducible and
high-resolution imaging technique that uses near-infrared light to generate 2D or 3D
images of tissues. Spectral-domain OCT (SD-OCT) is the fourth generation iteration
of this technology, which offers higher-definition and more rapid imaging, as
compared to previously available time-domain OCT (TD-OCT). SD-OCT allows
segmentation of discrete retinal layers within the macula, with each layer potentially
having different biological and clinical relevance in the context of neurological
diseases (figure 5.1). As the retina is unmyelinated, OCT offers a unique avenue to
examine neurodegenerative and inflammatory processes occurring outside of the
context of demyelination. Insights into these pathobiological processes are of
paramount importance, since neuroaxonal dysfunction and degeneration (rather
than demyelination) are now believed to be the principal susbstrates of clinical
disability in MS and related diseases [28–30]. Thus, OCT has become an extremely
useful imaging tool in these disease states, with a multitude of research establishing
the role of OCT-derived retinal layer thicknesses in the evaluation and monitoring of
MS and related disorders over the last two decades. Insights into these diseases
gained through OCT form an important context for more recent studies using
OCTA.

Figure 5.1. Example of a macular SD-OCT image. This macular image was acquired from a consenting
healthy research participant using a Cirrus SD-OCT device (model 5000, software version 8.1, Carl Zeiss
Meditec, California, USA), with automated segmentation of discrete retinal layers. RNFL = retinal nerve fiber
layer, GCL = ganglion cell layer, IPL = inner plexiform layer, INL = inner nuclear layer, OPL = outer
plexiform layer, ONL = outer nuclear layer, ELM = external limiting membrane. (Reprinted from [46], with
permission.)

5-3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

5.2.1 Retinal OCT in multiple sclerosis (MS)

Extensive studies using OCT in people with MS have been undertaken in recent
years. Peripapillary RNFL (pRNFL) and ganglion cell + inner plexiform layer
(GCIPL) thicknesses have been repeatedly shown to be reduced in the eyes of people
with MS as compared to healthy controls, and reductions are most marked in MS
eyes with a history of ON (MS-ON eyes) [31, 32]. GCIPL thickness measures are
more reliable and reproducible (with an inter-visit intra-class correlation coefficient
of 0.99 in both MS patients and healthy controls using Cirrus HD-OCT) [33] than
pRNFL thickness measures, and are less susceptible to confounding by edema in the
context of ON [33, 34]. In people with MS, GCIPL thickness has been demonstrated
to correlate with brain volume (particularly gray matter volumes), quantitative
spinal cord MRI measures, high and low-contrast visual function, and global
disability [33, 35–39]. Longitudinal studies have demonstrated that progressive
GCIPL atrophy occurs in MS, and correlates with worsening visual function, global
disability accumulation, and brain (and brain substructure) atrophy [40, 41].
Furthermore, accelerated GCIPL atrophy is seen in patients with clinical and/or
radiological inflammatory activity, and atrophy is differentially modulated by
various disease modifying therapies (DMT) [41, 42]. Based on these abundant
data, GCIPL thickness is thought to be a useful biomarker of neurodegeneration in
MS. In addition, other qualitative and quantitative OCT findings may also be
informative in MS; microcystic macular edema is associated with disability
progression, inner nuclear layer (INL) thickening is associated with inflammatory
disease activity, and thinning of the deeper retinal layers (the INL and the outer
nuclear layer—ONL) may be a hallmark of progressive MS [37, 43–45]. Overall,
OCT is considered to offer a ‘window into the brain’ in people with MS, and has
clinical applications as a biomarker for prognostication, monitoring of disease
progression, and evaluation of treatment response [46–48].

5.2.2 Retinal OCT in neuromyelitis optica spectrum disorder (NMOSD)

ON behaves somewhat differently in NMOSD as compared to MS. In patients with
aquaporin-4 seropositive NMOSD, ON is typically more severe, more frequently
bilateral, associated with poorer visual function outcomes, and more likely to be
recurrent [49–51]. This difference in severity is reflected in OCT measures, as reductions
in pRNFL and GCIPL thicknesses are substantially greater in NMOSD-ON eyes as
compared to MS-ON eyes, or idiopathic-ON eyes [49, 50, 52]. Additionally, micro-
cystic macular edema and INL thickening may also be present in NMOSD-ON eyes—
perhaps more frequently after severe ON episodes associated with worse structural
and functional outcomes [53, 54]. Similar to findings in MS, pRNFL and GCIPL
thickness correlate well with visual function measures in NMOSD-ON [50, 55, 56].
However, in contrast with MS, correlations between retinal layer thicknesses and
global disability measures such as the expanded disability status scale (EDSS) have
not been robustly and reproducibly demonstrated in NMOSD [49]. Similarly, the
evidence for subclinical optic neuropathy in NMOSD is not as conclusive as it is in
MS, although a number of studies suggest that reductions in pRNFL and GCIPL

5-4
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

thicknesses may be present in NMOSD eyes without a history of ON [52, 54, 57].
In addition, foveal thinning has been observed in NMOSD eyes with and without a
history of AON, as compared to control eyes, and thus has been suggested as an
alternative putative marker of subclinical optic neuropathy in these patients [58].
The lack of conclusive evidence for subclinical optic neuropathy or for a clear
relationship between retinal layer thicknesses and global disability measures in
NMOSD may be explained by the pathophysiology of the disorder. In NMOSD,
accumulation of disability is thought to be primarily attack-related, rather than
driven by subclinical pathophysiological processes [59]. By contrast, subclinical
inflammation and neuroaxonal loss are important components of MS disease
pathology (that are capitulated in the anterior visual pathway) [60], and much
disability accumulation in MS is not associated with overt attacks [61]. Furthermore,
large volume retinal tissue loss occurring after severe ON attacks in NMOSD may
mask potential relationships between retinal layer thicknesses and global disease
parameters in those eyes, and may also confound longitudinal changes through a
‘floor effect’ whereby further neuroaxonal loss may not be detectable.

5.3 Optical coherence tomography angiography (OCTA) of

the retina
OCTA is a relatively recent advancement of OCT technology, which detects
inherent motion of predominantly red blood cells in retinal vessels to generate
en face images of the retinal vasculature. OCTA imaging is a rapid and well-
tolerated technique, and does not require the injection of dye. Furthermore, the
ability of OCTA to produce depth-resolved images allowing segmentation of
discrete vascular plexuses represents a distinct advantage as compared to other
retinal vascular imaging techniques (such as fluorescein angiography).
The anatomy of the retinal vasculature is highly organized, with certain vascular
plexuses consistently lying within certain histological layers, and minimal variation
in this anatomy occurring between individuals [62]. Thus, retinal layer boundaries
detected using OCT can be used to segment OCTA images into distinct vascular
plexuses (figure 5.2). Arterial supply to the inner retina is derived from the central
retinal artery—a branch of the ophthalmic artery. The vascular plexuses supplying
the inner retinal layers include the radial peripapillary capillary plexus (RPCP), the
superficial vascular plexus (SVP), and the deep vascular plexus (DVP). Each of these
vascular plexuses comprises a network of arteries, arterioles, capillaries, venules and
veins. The SVP lies primarily within the ganglion cell layer and supplies the RNFL
and the GCIPL. The DVP lies primarily within the INL, and supplies the INL and
the outer plexiform layer (OPL). Arterial supply and venous drainage of the DVP is
through the SVP. The choriocapillaris lies within Bruch’s membrane and provides
vascular supply to the outer retina. Arterial supply to the choriocapillaris is derived
from the posterior ciliary arteries—a branch of the ophthalmic artery. Visualization
of the outer retinal vasculature with OCTA is considered by some experts to be
substantially more challenging than the inner retinal circulation, due to (1) projection

5-5
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 5.2. Examples of retinal OCTA images of the optic nerve head and macula. 3 mm × 3 mm angiography
images acquired from the left eye of a healthy control acquired using Cirrus AngioPlex SD-OCT (Carl Zeiss)
are demonstrated here, centered over the optic nerve head or fovea. Depth-encoded images are used to
generate images of the discrete vascular plexuses. A = optic nerve head depth-encoded image, B = optic nerve
head superficial vascular plexus, C = optic nerve head deep vascular plexus, D = optic nerve head
choriocapillaris, E = macular depth-encoded image, F = macular superficial vascular plexus with the foveal
avascular zone illustrated by the red circle, G = macular deep vascular plexus, H = macular choriocapillaris.
The foveal avascular zone is the capillary-free area at the center of macular images E–G.

artifact from the inner retinal vasculature, and (2) the hyper-reflective retinal pigment
epithelium (RPE) acting as a barrier to light beam penetration [63].
OCTA images can be reviewed qualitatively, which is particularly useful in the
clinical evaluation of certain ophthalmologic disorders associated with focal or
regional retinal abnormalities e.g. retinal artery or vein occlusions, and diabetic
retinopathy. In the case of neurological disorders that are thought to be associated
with a more diffuse effect on the retinal vasculature, quantitative OCTA measure-
ments are likely to be more useful than qualitative analyses. A range of different
quantitative measures have been derived using OCTA to date, including vascular
plexus density (sometimes referred to as vessel density), vessel length density, area of
the foveal avascular zone, flow index, flow velocity, skeleton density, fractal
dimension and non-perfused area [64]. In addition, a range of different techniques
(using various algorithms) for the derivation of these measurements has been
reported. The choice of quantitative metrics analyzed may vary according to the
device utilized, the disease-state being examined, and the study hypothesis.

5.3.1 OCTA in MS
Retinal OCTA has become a research tool of interest in people with MS in recent
years. A number of studies have explored retinal OCTA measures in people with
MS, and compared these to healthy controls (table 5.1).
Macular SVP density has repeatedly been demonstrated to be lower in people
with MS than in healthy controls [67, 70, 71]. Correlations between vascular plexus

5-6
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 5.1. Selection of published studies exploring OCTA measures in people with MS.

Study authors Study design OCTA measures Key findings

Wang et al Cross-sectional study ONH-FI and • ONH-FI was lower in MS-ON

2014 [65] of 35 people with macular eyes as compared to healthy con-
MS and 21 healthy parafoveal FI trol eyes and MS-NON eyes
controls • No significant differences in par-
afoveal FI between MS-ON/MS-
NON eyes and healthy control
eyes

Spain et al Cross-sectional study ONH-FI • ONH-FI was lower in MS-ON

2018 [66] of 45 people with and MS-NON eyes as compared
MS and 32 healthy to healthy control eyes
controls • No significant correlations
between ONH-FI and retinal
layer thicknesses

Lanzillo et al Cross-sectional study Macular SVP • SVP density was lower in MS-
2017 [67] of 50 people with density ON eyes and MS-NON eyes as
MS and 46 healthy compared to healthy control eyes
controls • SVP density significantly corre-
lated with RNFL and GCC
thicknesses
• Lower SVP density significantly
correlated with higher disability
(EDSS and MSSS scores)

Feucht et al Cross-sectional study Macular SVP, • SVP and DVP densities were
2018 [68] of 42 people with DVP and lower in MS-ON eyes as com-
MS and 50 healthy choriocapillaris pared to MS-NON eyes and
controls densities healthy control eyes
• SVP and DVP densities signifi-
cantly correlated with GCIPL
volumes, INL volumes and TMV

Lanzillo et al Longitudinal study Macular SVP • Across the whole group, SVP
2018 [69] (1 year) of 50 density density in the parafoveal region
people with MS showed a small but significant
increase over 1 year
• Reductions in SVP density in the
parafoveal region correlated sig-
nificantly with increasing disabil-
ity (EDSS scores)

(Continued)

5-7
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 5.1. (Continued )

Study authors Study design OCTA measures Key findings

Murphy et al Cross-sectional study Macular SVP and • SVP density was lower in MS-
2019 [70] of 111 people with DVP densities ON and MS-NON eyes as com-
MS and 50 healthy pared to healthy control eyes
controls • Lower SVP density correlated
significantly with longer disease
duration, higher disability (EDSS
and MSFC scores), and poorer
high- and low-contrast letter
acuity scores
• SVP density correlated signifi-
cantly with RNFL and GCIPL
thicknesses, DVP density corre-
lated significantly with GCIPL
and INL thicknesses

Yilmaz et al Cross-sectional study Macular SVP and • SVP and DVP densities were
2020 [71] of 47 people with DVP densities, lower in people with MS as com-
MS and 61 healthy RPCP density, pared to healthy controls, across
controls FAZ area all EDTRS sectors of the macula
• No significant differences in FAZ
and RPCP density between peo-
ple with MS and healthy controls
• SVP and DVP densities were
inversely correlated with visual
evoked potential (VEP) latency

ONH-FI = optic nerve head flow index, FI = flow index, MS-ON = MS optic neuritis, MS-NON = MS non-
optic neuritis, SVP = superficial vascular plexus, DVP = deep vascular plexus, RNFL = retinal nerve fiber
layer, GCC = ganglion cell complex, GCIPL = ganglion cell + inner plexiform layer, INL = inner nuclear
layer, TMV = total macular volume, EDSS = expanded disability status scale, MSSS = multiple sclerosis
severity scale, MSFC = multiple sclerosis functional composite, RPCP = radial peripapillary capillary plexus,
FAZ = foveal avascular zone.

densities and retinal layer thicknesses have been observed in both MS eyes and
healthy control eyes, suggesting a strong relationship between structural and
vascular measures in the macula [67, 68, 70, 71]. Additionally, SVP density
reductions in MS eyes as compared to healthy control eyes have been demonstrated
to be present in all sectors of the EDTRS grid, demonstrating that alterations in
vascular measures occur diffusely within the macula in MS [67, 71]. Reductions
in SVP density appear to be more marked in MS-ON eyes than in MS non-ON
(MS-NON) eyes, however significant differences have been detected even in
MS-NON eyes alone as compared to healthy control eyes [67, 70]. In some cases,
the rarefication of the macular SVP in MS-ON eyes may be so evident as to be visible
to the naked eye with qualitative image evaluation (figure 5.3). It is notable that one

5-8
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 5.3. Macular SVP OCTA image in the MS-NON eye and MS-ON eye of one individual with MS.
These images show the qualitative differences in the SVP visible on macular Spectralis OCTA (Heidelberg
Engineering, Germany) images from the eyes of an individual with MS who has a history of unilateral ON.
The patient’s left eye (A) was unaffected by ON and had an SVP density of 26.9% calculated using an image
binarization method, whereas the patient’s right eye (B) was affected by ON 5 years previously and had an
SVP density of 11.7% calculated using the same method. The capillary network is noticeably less dense in the
eye with a history of ON.

study failed to identify significant differences in macular SVP density of MS-NON

eyes versus healthy control eyes, although that study was relatively small and average
MS disease duration was short [68]. As the magnitude of SVP reductions in MS-NON
appears to be lower than in MS-ON eyes, it is likely that examining larger patient
cohorts with longer disease durations is necessary to detect macular SVP density
reductions in MS-NON eyes.
Differences in OCTA measures between people with MS and healthy controls
have also been identified at the optic nerve head. Specifically, optic nerve head flow
index (ONH-FI) has been shown to be lower in MS-ON eyes as compared to healthy
control eyes [65, 66]. Additionally, one study suggested that ONH-FI may be lower
in MS-NON eyes as compared to healthy control eyes [66]. Two studies have also
compared OCTA measures at the optic disc in ON eyes to non-arteritic ischemic
optic neuropathy (NAION) eyes, with both studies finding no difference in RPCP
density between these groups at either the acute or chronic stages [72, 73].
Based on data published to date, it appears that OCTA-derived macular vascular
measures may demonstrate relationships with certain clinical parameters in MS.
Studies have shown that lower SVP density is associated with (1) longer disease
duration, (2) higher levels of global disability measured by the expanded disability
status scale (EDSS), multiple sclerosis severity scale (MSSS) and multiple sclerosis
functional composite (MSFC) scores, (3) poorer visual function measured by high-
and low-contrast letter acuity scores, and (4) longer visual evoked potential (VEP)
latency [67, 70, 71]. Furthermore, a short longitudinal study (1 year of follow-up)
identified that progressive reductions in SVP density were significantly associated
with increasing levels of disability [69]. Studies have noted the importance of
considering ON history in study methodology and statistical analyses, since the
magnitude of retinal tissue loss that occurs after ON is known to confound the

5-9
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

relationships between retinal layer thicknesses and clinical characteristics in MS [74],

and it is assumed that similar confounding occurs with vessel density reductions after
ON. Once this confounding is controlled for, there is certainly accumulating
evidence that OCTA-derived vascular measures have the potential to be a useful
biomarker of disease severity in MS.

5.3.2 OCTA in NMOSD

OCTA measures have not been as well-explored in NMOSD to date as in MS,
although a number of studies have already provided interesting results. Lower
RPCP and SVP densities have been identified in people with NMOSD as compared
to healthy controls, with differences compared to healthy controls being more
marked in NMOSD eyes with a history of ON (NMOSD-ON eyes) than in NMOSD
eyes without a history of ON (NMOSD-NON eyes) [75–77]. Lower DVP density has
also been detected in some people with NMOSD as compared to healthy controls
[77]. People with NMOSD often experience multiple attacks of ON, and a higher
number of prior ON attacks is associated with lower SVP and DVP densities in those
eyes [77].
Similar to relationships identified in MS eyes, vascular plexus densities in
NMOSD eyes have been found to correlate well with retinal layer thicknesses
[76, 77]. However, in one study where RPCP and SVP densities were found to be
lower in NMOSD-NON eyes than healthy control eyes, pRNFL thickness was not
significantly different between the same groups, suggesting that vascular densities
can be altered in these patients even where substantial axonal loss is not detected
[76]. RPCP, SVP and DVP have all been suggested to correlate with visual function
in patients with NMOSD [76, 77].

5.4 Limitations of OCTA

OCTA images can be acquired using a range of commercially available devices,
which employ different algorithms to segment the retinal vascular plexuses accord-
ing to retinal layer boundaries. The quantitative vascular measurements which can
be acquired (vascular plexus densities, flow indices, foveal avascular zone area) may
differ across devices. Accurate segmentation of the vascular plexus layers may be
impaired by the presence of pathology distorting the retinal tissue (such as cysts)
[78]. The number of vascular structures which can be segmented and the boundaries
used for these may also differ, e.g. the macular DVP can be subdivided into the
intermediate capillary plexus and the deep capillary plexus, and the RPCP can be
combined with the SVP [62]. In addition, there is a lack of accepted standardized
nomenclature in the field of OCTA, with the terms ‘capillary plexus’, ‘vascular
plexus’ and ‘vascular complex’ often used interchangeably to describe vascular
structures, and the terms ‘vessel density’, ‘vascular plexus density’ and ‘vessel density
index’ variably used to report derived measurements. Furthermore, some devices
provide images of the deeper vascular structures (such as the choriocapillaris and
choroid) while other devices do not. Visualizing the vasculature beyond the highly
reflective RPE using OCTA is technically challenging, and where devices do provide

5-10
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

images of these deeper vascular structures, there remains some uncertainty regarding
the reliability of these images and the quantitative measures gleaned from them
[63, 79]. In addition to differences across devices, many research studies employing
OCTA utilize different image processing tools or novel research algorithms, adding
further complexity to this field. For the reasons outlined here, it is not clear yet
whether quantitative data acquired from different devices or processed using
difference algorithms can be directly compared, and accordingly calls into question
the generalizability of study findings.
Artifacts are known to be a significant issue impacting the acquisition and
interpretation of OCTA images, across healthy controls and various disease states
[80–82]. Possible artifacts include projection artifact (from inner vascular structures
onto deeper structures), motion artifact, loss of focus artifact (figure 5.4), and blink
lines, amongst others [82]. It has been demonstrated that imaging artifacts are more
frequent in people with MS as compared to healthy controls, and that these artifacts
may lead to underestimation of vascular plexus densities in people with MS [82]. The
greater frequency of imaging artifact in people with MS may be due to nystagmus,
poor visual acuity, fatigue, and physical disability—factors which may be also be
present in people with NMOSD or other disabling neurological diseases. However,
many research studies employing OCTA fail to describe clear quality control
mechanisms to counteract the difficulties associated with either image acquisition
or artifact-degradation.
There is still much to be understood regarding the interpretation of OCTA
measurements. Even in healthy controls, OCTA measures appear to be influenced
by demographic factors, including age and ethnicity [83, 84], yet there is a lack of
accepted age and ethnicity-adjusted normative data available for clinical, or research
application. OCTA images are typically generated and quantitatively analyzed in
en face format, and thus are usually not normalized or adjusted for tissue volume,

Figure 5.4. Impact of artifact on visualization of the SVP with OCTA, in a single eye. These SVP images were
acquired from the same eye of the same subject during a single sitting, using Spectralis OCTA (Heidelberg
Engineering, Germany). Image A is degraded by loss of focus artifact, resulting in a lack of definition of the
capillary architecture, and apparent ‘dropout’ of the smallest vessels. By contrast, image B is a high-quality
image with minimal artifact. The obvious difference between these images illustrates how artifacts can impair
the qualitative and quantitative interpretation of OCTA images.

5-11
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

which may affect the interpretation of the vascular parameters. Moreover, the
histological correlate of altered vascular plexuses is uncertain in MS and NMOSD.
Since OCTA depends on vascular flow to generate images, reduced vascular plexus
densities can represent loss of capillaries, shrinking of capillaries, or simply slow flow
below the limits of motion-based detection. The pathobiological processes driving
vascular density alterations in demyelinating CNS disease could include reduced
metabolic demand due to retinal atrophy, altered metabolism due to neuronal
dysfunction, primary vascular processes, or a combination of these factors. Retinal
periphlebitis is also well-documented in MS [5, 21], and may impair retinal vascular
flow. Furthermore, metabolic and vascular diseases such as hypertension and
diabetes mellitus can impact the retinal vasculature, making interpretation of
findings complex in patients with co-morbidities.

5.5 Conclusions
OCTA has become a useful non-invasive tool to examine the retinal vasculature
in vivo. MS and NMOSD are neuroimmune disorders with prominent anterior
visual pathway involvement, and there is a growing body of research employing
OCTA in these disorders. Retinal vascular measurements appear to be reduced
across MS eyes, and may correlate with visual function and global disability, thus
representing a potential biomarker of disease. In NMOSD, research is more limited
but also suggests reduced retinal vascular measurements, particularly in eyes
previously affected by ON. OCTA technology continues to improve, yet there are
a number of limitations which should be considered—including technical limitations
and challenges in the interpretation of findings. Further research should focus on a
range of issues including the impact of demographic characteristics and co-morbid-
ities on OCTA measurements in individuals; the standardization of quality control
processes across research studies; and longitudinal studies examining changes in
retinal vascular measures over time in MS and NMOSD.

References
[1] Ikuta F and Zimmerman H M 1976 Distribution of plaques in seventy autopsy cases of
multiple sclerosis in the united states Neurology 26 26–8
[2] Toussaint D, Périer O, Verstappen A and Bervoets S 1983 Clinicopathological study of the
visual pathways, eyes, and cerebral hemispheres in 32 cases of disseminated sclerosis J. Clin.
Neuroophthalmol. 3 211–20
[3] Kitley J, Leite M I and Nakashima I et al 2012 Prognostic factors and disease course in
aquaporin-4 antibody-positive patients with neuromyelitis optica spectrum disorder from the
United Kingdom and Japan Brain: A J. Neurol. 135 1834–49
[4] Kim S, Kim W, Li X F, Jung I and Kim H J 2012 Clinical spectrum of CNS aquaporin-4
autoimmunity Neurology 78 1179–85
[5] Green A J, McQuaid S, Hauser S L, Allen I V and Lyness R 2010 Ocular pathology in
multiple sclerosis: retinal atrophy and inflammation irrespective of disease duration Brain
133 1591–601
[6] Kerrison J B, Flynn T and Green W R 1994 Retinal pathologic changes in multiple sclerosis
Retina (Philadelphia, Pa) 14 445–51

5-12
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[7] Minagar A, Jy W and Jimenez J J et al 2001 Elevated plasma endothelial microparticles in

multiple sclerosis Neurology 56 1319–24
[8] Aksungar F B, Topkaya A E, Yildiz Z, Sahin S and Turk U 2008 Coagulation status and
biochemical and inflammatory markers in multiple sclerosis J. Clin. Neurosci. 15 393–7
[9] Dardiotis E, Arseniou S and Sokratous M et al 2017 Vitamin B12, folate, and homocysteine
levels and multiple sclerosis: a meta-analysis Mult. Scler. Relat. Disord. 17 190–7
[10] Cramer S P, Modvig S, Simonsen H J, Frederiksen J L and Larsson H B W 2015
Permeability of the blood–brain barrier predicts conversion from optic neuritis to multiple
sclerosis Brain 138 2571–83
[11] Wakefield A J, More L J, Difford J and McLaughlin J E 1994 Immunohistochemical study
of vascular injury in acute multiple sclerosis J. Clin. Pathol. 47 129–33
[12] Adams C W 1988 Perivascular iron deposition and other vascular damage in multiple
sclerosis J. Neurol. Neurosurg. Psychiatry 51 260–5
[13] Popescu B F G and Lucchinetti C F 2016 Immunopathology: autoimmune glial diseases and
differentiation from multiple sclerosis Handb. Clin. Neurol. 133 95–106
[14] Saindane A M, Law M, Ge Y, Johnson G, Babb J S and Grossman R I 2007 Correlation of
diffusion tensor and dynamic perfusion MR imaging metrics in normal-appearing corpus
callosum: support for primary hypoperfusion in multiple sclerosis AJNR Am. J. Neuroradiol.
28 767–72
[15] Law M, Saindane A M and Ge Y et al 2004 Microvascular abnormality in relapsing-
remitting multiple sclerosis: perfusion MR imaging findings in normal-appearing white
matter Radiology 231 645–52
[16] Varga A W, Johnson G, Babb J S, Herbert J, Grossman R I and Inglese M 2009 White
matter hemodynamic abnormalities precede sub-cortical gray matter changes in multiple
sclerosis J. Neurol. Sci. 282 28–33
[17] Adhya S, Johnson G, Herbert J, Jaggi H, Babb J S, Grossman R I and Inglese M 2006
Pattern of hemodynamic impairment in multiple sclerosis: dynamic susceptibility contrast
perfusion MR imaging at 3.0 T Neuroimage 33 1029–35
[18] Zhang X, Guo X and Zhang N et al 2018 Cerebral blood flow changes in multiple sclerosis
and neuromyelitis optica and their correlations with clinical disability Front. Neurol. 9 305
[19] Maggi P, Absinta M and Grammatico M et al 2018 Central vein sign differentiates multiple
sclerosis from central nervous system inflammatory vasculopathies Ann. Neurol. 83 283–94
[20] Absinta M, Nair G, Sati P, Cortese I C M, Filippi M and Reich D S 2015 Direct MRI
detection of impending plaque development in multiple sclerosis Neurol. Neuroimmunol.
Neuroinflamm. 2 e145
[21] Rucker C W 1944 Sheathing of the retinal veins in multiple sclerosis Mayo. Clin. Proc. 19
176–8
[22] Hamann S, Zeuthen T, Cour M L, Nagelhus E A, Ottersen O P, Agre P and Nielsen S 1998
Aquaporins in complex tissues: distribution of aquaporins 1–5 in human and rat eye Am. J.
Physiol.: Cell Physiol. 274 C1332
[23] Zeka B, Hastermann M and Kaufmann N et al 2016 Aquaporin 4-specific T cells and NMO-
IgG cause primary retinal damage in experimental NMO/SD Acta Neuropathol. Commun. 4 82
[24] Younge B R 1976 Fluorescein angiography and retinal venous sheathing in multiple sclerosis
Can. J. Ophthalmol. 11 31–6
[25] Birch M K, Barbosa S, Blumhardt L D, O’Brien C and Harding S P 1996 Retinal venous
sheathing and the blood-retinal barrier in multiple sclerosis Arch. Ophthalmol. 114 34–9

5-13
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[26] Lightman S, McDonald W I, Bird A C, Francis D A, Hoskins A, Batchelor J R and Halliday

A M 1987 Retinal venous sheathing in optic neuritis. its significance for the pathogenesis of
multiple sclerosis Brain 110 405–14
[27] Green A J and Cree B A C 2009 Distinctive retinal nerve fibre layer and vascular changes in
neuromyelitis optica following optic neuritis J. Neurol. Neurosurg. Psychiatry 80 1002–5
[28] Dutta R and Trapp B D 2011 Mechanisms of neuronal dysfunction and degeneration in
multiple sclerosis Prog. Neurobiol. 93 1–12
[29] Correale J, Gaitán M I, Ysrraelit M C and Fiol M P 2017 Progressive multiple sclerosis:
from pathogenic mechanisms to treatment Brain 140 527–46
[30] Trapp B D, Peterson J, Ransohoff R M, Rudick R, Mörk S and Bö L 1998 Axonal
transection in the lesions of multiple sclerosis N. Engl. J. Med. 338 278–85
[31] Petzold A, Balcer L J and Calabresi P A et al 2017 Retinal layer segmentation in multiple
sclerosis: a systematic review and meta-analysis Lancet Neurol. 16 797–812
[32] Petzold A, de Boer J F and Schippling S et al 2010 Optical coherence tomography in multiple
sclerosis: a systematic review and meta-analysis The Lancet Neurol. 9 921–32
[33] Saidha S, Syc S B and Ibrahim M A et al 2011 Primary retinal pathology in multiple sclerosis
as detected by optical coherence tomography Brain 134 518–33
[34] Syc S B, Saidha S and Newsome S D et al 2012 Optical coherence tomography segmentation
reveals ganglion cell layer pathology after optic neuritis Brain 135 521–33
[35] Nguyen J, Rothman A and Gonzalez N et al 2019 Macular ganglion cell and inner plexiform
layer thickness is more strongly associated with visual function in multiple sclerosis than
bruch membrane opening-minimum rim width or peripapillary retinal nerve fiber layer
thicknesses J. Neuroophthalmol.
[36] Nguyen J, Rothman A and Fitzgerald K et al 2018 Visual pathway measures are associated
with neuropsychological function in multiple sclerosis Curr. Eye Res. 43 941–8
[37] Saidha S, Sotirchos E S and Oh J et al 2013 Relationships between retinal axonal and
neuronal measures and global central nervous system pathology in multiple sclerosis JAMA
Neurol 70 34–43
[38] Saidha S, Syc S B and Durbin M K et al 2011 Visual dysfunction in multiple sclerosis correlates
better with optical coherence tomography derived estimates of macular ganglion cell layer
thickness than peripapillary retinal nerve fiber layer thickness Mult. Scler. 17 1449–63
[39] Oh J, Sotirchos E S and Saidha S et al 2015 Relationships between quantitative spinal cord
MRI and retinal layers in multiple sclerosis Neurology 84 720–8
[40] Saidha S, Al-Louzi O and Ratchford J N et al 2015 Optical coherence tomography reflects
brain atrophy in multiple sclerosis: a four-year study Ann. Neurol. 78 801–13
[41] Ratchford J N, Saidha S and Sotirchos E S et al 2013 Active MS is associated with
accelerated retinal ganglion cell/inner plexiform layer thinning Neurology 80 47–54
[42] Button J, Al-Louzi O and Lang A et al 2017 Disease-modifying therapies modulate retinal
atrophy in multiple sclerosis: a retrospective study Neurology 88 525–32
[43] Gelfand J M, Nolan R, Schwartz D M, Graves J and Green A J 2012 Microcystic macular
oedema in multiple sclerosis is associated with disease severity Brain: A J. Neurol. 135 1786–93
[44] Saidha S, Sotirchos E S and Ibrahim M A et al 2012 Microcystic macular oedema, thickness
of the inner nuclear layer of the retina, and disease characteristics in multiple sclerosis: a
retrospective study Lancet Neurol. 11 963–72
[45] Sotirchos E S, Caldito N G and Filippatou A et al 2020 Progressive multiple sclerosis is
associated with faster and specific retinal layer atrophy Ann. Neurol.

5-14
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[46] Lambe J, Murphy O and Saidha S 2018 Can optical coherence tomography be used to guide
treatment decisions in adult or pediatric multiple sclerosis? Curr. Treat. Options Neurol. 20
1–15
[47] Sotirchos E S and Saidha S 2018 OCT is an alternative to MRI for monitoring MS—YES
Mult. Scler. 24 701–3
[48] Lambe J, Saidha S and Bermel R A 2020 Optical coherence tomography and multiple
sclerosis: update on clinical application and role in clinical trials Mult. Scler. 26 624–39
[49] Green A, Bennett J and Lana-Peixoto M et al 2015 Neuromyelitis optica and multiple
sclerosis: seeing differences through optical coherence tomography Multiple Sclerosis J. 21
678–88
[50] Sotirchos E S, Filippatou A and Fitzgerald K C et al 2019 Aquaporin-4 IgG seropositivity is
associated with worse visual outcomes after optic neuritis than MOG-IgG seropositivity and
multiple sclerosis, independent of macular ganglion cell layer thinning Mult. Scler. J. 26
1360–71
[51] Fernandes D, Ramos R I, Falcochio C, Apóstolos-Pereira S, Callegaro D and Monteiro M
L 2012 Comparison of visual acuity and automated perimetry findings in patients with
neuromyelitis optica or multiple sclerosis after single or multiple attacks of optic neuritis J.
Neuro-Ophthalmol. 32 102–6
[52] Peng A, Qiu X, Zhang L, Zhu X, He S, Lai W and Chen L 2017 Evaluation of the retinal
nerve fiber layer in neuromyelitis optica spectrum disorders: a systematic review and meta-
analysis J. Neurol. Sci. 383 108–13
[53] Kaufhold F, Zimmermann H, Schneider E, Ruprecht K, Paul F, Oberwahrenbrock T and
Brandt A U 2013 Optic neuritis is associated with inner nuclear layer thickening and
microcystic macular edema independently of multiple sclerosis PLoS One 8 e71145
[54] Sotirchos E S, Saidha S and Byraiah G et al 2013 In vivo identification of morphologic
retinal abnormalities in neuromyelitis optica Neurology 80 1406–14
[55] Cheng L, Wang J, He X, Xu X and Ling Z F 2016 Macular changes of neuromyelitis optica
through spectral-domain optical coherence tomography Int. J. Ophthalmol. 9 1638–45
[56] Peng C, Wang W and Xu Q et al 2016 Structural alterations of segmented macular inner
layers in Aquaporin4-antibody-positive optic neuritis patients in a chinese population PLoS
One 11 e0157645
[57] Hu S J and Lu P R 2018 Retinal ganglion cell-inner plexiform and nerve fiber layers in
neuromyelitis optica Int. J. Ophthalmol. 11 89–93
[58] Jeong I, Kim H, Kim N, Jeong K and Park C 2016 Subclinical primary retinal pathology in
neuromyelitis optica spectrum disorder J. Neurol. 263 1343–8
[59] Pittock S J and Lucchinetti C F 2016 Neuromyelitis optica and the evolving spectrum of
autoimmune aquaporin-4 channelopathies: a decade later Ann. N. Y. Acad. Sci. 1366 20–39
[60] Reich D S, Lucchinetti C F and Calabresi P A 2018 Multiple sclerosis N. Engl. J. Med. 378
169–80
[61] Kappos L, Wolinsky J S and Giovannoni G et al 2020 Contribution of relapse-independent
progression vs relapse-associated worsening to overall confirmed disability accumulation in
typical relapsing multiple sclerosis in a pooled analysis of 2 randomized clinical trials JAMA
Neurol. 77 1132–40
[62] Campbell J P, Zhang M, Hwang T S, Bailey S T, Wilson D J, Jia Y and Huang D 2017
Detailed vascular anatomy of the human retina by projection-resolved optical coherence
tomography angiography Sci. Rep. 7 42201

5-15
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[63] Wang J C, Laíns I, Silverman R F, Sobrin L, Vavvas D G, Miller J W and Miller J B 2018
Visualization of choriocapillaris and choroidal vasculature in healthy eyes with en face swept-
source optical coherence tomography versus angiography Transl. Vis. Sci. Technol. 7 25
[64] Coffey A M, Hutton E K, Combe L, Bhindi P, Gertig D and Constable P A 2020 Optical
coherence tomography angiography in primary eye care Clin. Exp. Optom. 104 3–13
[65] Wang X, Jia Y and Spain R et al 2014 Optical coherence tomography angiography of optic
nerve head and parafovea in multiple sclerosis Br. J. Ophthalmol. 98 1368–73
[66] Spain R I, Liu L, Zhang X, Jia Y, Tan O, Bourdette D and Huang D 2018 Optical coherence
tomography angiography enhances the detection of optic nerve damage in multiple sclerosis
Br. J. Ophthalmol. 102 520–4
[67] Lanzillo R, Cennamo G and Criscuolo C et al 2017 Optical coherence tomography
angiography retinal vascular network assessment in multiple sclerosis Mult. Scler. 24
1706–14
[68] Feucht N, Maier M and Lepennetier G et al 2018 Optical coherence tomography
angiography indicates associations of the retinal vascular network and disease activity in
multiple sclerosis Mult. Scler. 25 224–34
[69] Lanzillo R, Cennamo G and Moccia M et al 2018 Retinal vascular density in multiple
sclerosis: a one year follow up Eur. J. Neurol. 26 198–201
[70] Murphy O C, Kwakyi O and Iftikhar M et al 2019 Alterations in the retinal vasculature
occur in multiple sclerosis and exhibit novel correlations with disability and visual function
measures Mult. Scler. 26 815–28
[71] Yilmaz H, Ersoy A and Icel E 2020 Assessments of vessel density and foveal avascular zone
metrics in multiple sclerosis: an optical coherence tomography angiography study Eye
(Lond.) 34 771–8
[72] Fard M A, Jalili J, Sahraiyan A, Khojasteh H, Hejazi M, Ritch R and Subramanian P S
2018 Optical coherence tomography angiography in optic disc swelling Am. J. Ophthalmol.
191 116–23
[73] Fard M, Yadegari S, Ghahvechian H, Moghimi S, Soltani-Moghaddam R and Subramanian
P 2019 Optical coherence tomography angiography of a pale optic disc in demyelinating
optic neuritis and ischemic optic neuropathy J. Neuro-Ophthalmol. 39 339–44
[74] Martinez-Lapiscina E H, Arnow S and Wilson J A et al 2016 Retinal thickness measured
with optical coherence tomography and risk of disability worsening in multiple sclerosis: a
cohort study Lancet Neurol 15 574–84
[75] Chen Y, Shi C, Zhou L, Huang S, Shen M and He Z 2020 The detection of retina
microvascular density in subclinical aquaporin-4 antibody seropositive neuromyelitis optica
spectrum disorders Front. Neurol. 11 35
[76] Huang Y, Zhou L and ZhangBao J et al 2019 Peripapillary and parafoveal vascular network
assessment by optical coherence tomography angiography in aquaporin-4 antibody-positive
neuromyelitis optica spectrum disorders Br. J. Ophthalmol. 103 789–96
[77] Kwapong W R, Peng C, He Z, Zhuang X, Shen M and Lu F 2018 Altered macular
microvasculature in neuromyelitis optica spectrum disorders Am. J. Ophthalmol. 192 47–55
[78] Wang J C and Miller J B 2019 Optical coherence tomography angiography: review of current
technical aspects and applications in chorioretinal disease Semin. Ophthalmol. 34 211–7
[79] Diaz J D, Wang J C and Oellers P et al 2018 Imaging the deep choroidal vasculature using
spectral domain and swept source optical coherence tomography angiography J. Vitreoretin.
Dis. 2 146–54

5-16
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[80] Spaide R F, Fujimoto J G and Waheed N K 2015 Image artifacts in optical coherence
tomography Retina 35 2163–80
[81] Enders C, Lang G E, Dreyhaupt J, Loidl M, Lang G K and Werner J U 2019 Quantity and
quality of image artifacts in optical coherence tomography angiography PLoS One 14
e0210505
[82] Iftikhar M, Zafar S and Gonzalez N et al 2019 Image artifacts in optical coherence
tomography angiography among patients with multiple sclerosis Curr. Eye Res. 44 558–63
[83] Zheng F, Zhang Q and Shi Y et al 2019 Age-dependent changes in the macular
choriocapillaris of normal eyes imaged with swept-source optical coherence tomography
angiography Am. J. Ophthalmol. 200 110–22
[84] Chun L Y, Silas M R, Dimitroyannis R C, Ho K and Skondra D 2019 Differences in
macular capillary parameters between healthy black and white subjects with optical
coherence tomography angiography (OCTA) PLoS One 14 e0223142

5-17
 IOP Publishing

Photo Acoustic and Optical Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
Ayman El-Baz and Jasjit S Suri

Chapter 6
Optical coherence tomography angiography for
the diagnosis of polypoidal choroidal
vasculopathy
Talisa De Carlo and Gregg T Kokame

Optical coherence tomography angiography (OCTA) is a novel non-invasive

technology that can be useful for detecting and monitoring polypoidal choroidal
vasculopathy (PCV) lesions. It is important to distinguish this subtype from other
forms of age-related macular degeneration because the treatment algorithm may be
different and it predicts anti-vascular endothelial growth factor resistance in all
ethnic populations. We describe how to detect PCV using both the OCT angiograms
and corresponding b-scans.

6.1 Introduction
Optical coherence tomography angiography (OCTA) is a relatively novel approach
to visualizing the retinal and choroidal vasculature in vivo. The technique uses non-
invasive motion-contrast to generate volumetric angiographic images in less than
one minute. OCTA compares the differences in backscattered OCT signal intensity,
or decorrelation signal, between sequential OCT b-scans obtained at each given
cross-section of the retina. The decorrelation signal between repeated OCT b-scans
are assumed to correlate to erythrocyte movement within the retinal vasculature and
can therefore be utilized to construct a blood flow map of the retina. OCTA images
correlate well with fluorescein angiography (FA) and indocyanine green angiog-
raphy (ICGA), providing even more detailed and high-resolution images of the
retinal and choroidal microvasculature. Furthermore, the OCTA images can be
segmented to show the individual vascular plexuses.
Each OCTA image set contains en face OCT angiograms, OCT b-scans, and
structural en face OCT images that are co-registered in order to visualize both

doi:10.1088/978-0-7503-2060-3ch6 6-1 ª IOP Publishing Ltd 2021

 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

structural and blood flow information in tandem. This allows for colocalization of
microvascular pathology to the structural images. Using the entire OCTA image set
or multiple components offers more information than just the OCT angiogram or
OCT b-scans alone. This can be particularly helpful in diseases such as polypoidal
choroidal vasculopathy (PCV), as we will see in this chapter.
PCV is a subtype of exudative or wet age-related macular degeneration (AMD),
in which polypoidal-shaped dilated lesions are a prominent part of the choroidal
neovascular network. This subtype has a different prevalence in different ethnic
populations [1, 2]. In Asian populations PCV can often make up over 50% of
patients presenting with exudative AMD. Caucasian patients were initially felt to
have a low incidence of PCV less than 10%, but with better screening with ICGA
using the scanning laser ophthalmoscope, prevalence has now been shown to be
20%–31% in Caucasian populations. Although wet AMD is infrequent in the black
population, when it occurs, it is usually associated with PCV.
PCV is the most important subtype of exudative AMD to identify, because it
predicts for anti-VEGF resistance [3]. PCV eyes are more anti-VEGF resistant than
typical wet AMD eyes, and this is critical, as anti-VEGF medications, such as
bevacizumab, ranibizumab, aflibercept, and brolucizumab, are the first line treat-
ment for exudative AMD. In the recent EVEREST II study, it was shown that
combination therapy with verteporfin photodynamic therapy (vPDT) not only
resulted in better vision than anti-VEGF monotherapy, but this was accomplished
with one half the number of intravitreal injections compared to anti-VEGF
monotherapy [4]. In addition, PCV responds different to different anti-VEGF
agents with a better response anatomically with brolucizumab, which is better
than aflibercept, which is better than ranibizumab [5].

6.2 OCTA for the diagnosis of PCV

OCTA is a promising and continually developing new technology for noninvasive
imaging of choroidal neovascularization related to exudative AMD. For the
diagnosis of PCV OCTA is not as sensitive as the gold standard of ICGA, as the
diagnostic polypoidal lesions are often not seen on OCTA, possibly due to slower
blood flow through the polypoidal lesions. However, the branching vascular
network (BVN) is often very well imaged with OCTA, again possibly due to faster
blood flow through the BVN. Because of the non-invasive nature of OCTA, it may
be very useful in following the response to therapy of PCV anatomically.

6.2.1 OCT angiograms

The PCV complex can be visualized on the OCT angiogram as a branching vascular
network (BVN) and circular flow lesions representing polyps [6]. The polyps can be
seen as either high-flow vascular dilations or faint low-flow lesions (figure 6.1). PCV
is a type 1 choroidal neovascular complex, located between Bruch’s membrane and
the retinal pigment epithelium. Therefore, segmentation of the OCT angiogram to
this region or to the choriocapillaris can give a higher yield for detection of the PCV
complex.

6-2
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 6.1. Two examples of optical coherence tomography angiograms of polypoidal choroidal vasculopathy
complexes. The polyps (arrowheads) can be seen as either vascular dilations (left) or faint low-flow lesions
(right) adjacent to branching vascular networks (arrows). This is seen best on the outer retina (left) or
choriocapillaris (right) segmentations.

Figure 6.2. The branching vascular network (arrows) of the polypoidal choroidal vasculopathy complex is
readily seen on the optical coherence tomography angiograms but no polyps are definitively visualized
although the rounded branches at the upper left aspect of the complex are suspicious for possible polyps. The
corresponding optical coherence tomography b-scans are invaluable in equivocal cases such as these.

The BVN is readily delineated on the OCT angiogram but the polyps are
visualized less consistently ranging from approximately 75%–92% of the time
(figure 6.2) [6–10]. Detection of the polyps is limited by a couple of factors.
Polyps that are too small or have too slow flow can be missed on the OCT
angiogram because of the principle of slowest detectable flow determined by the time
between repeated OCT b-scans [11]. If the blood flow within a polyp is slower than

6-3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

the time between repeated OCT b-scans, then no decorrelation signal will be
produced, resulting in no blood flow on the OCT angiogram. One study compared
characteristics of visualized versus non-visualized polyps on OCTA noting that
polyps with a higher height, that have a pulsatile nature, or that have overlying
hemorrhage may be less likely to be visualized on OCTA [12].
Furthermore, in eyes that have undergone treatment the polyps may regress while
the BVN remains. Therefore, detecting the entire PCV complex can be difficult in
previously treated eyes. Studies comparing OCTA to ICGA in treatment naïve eyes
demonstrated that the imaging modalities were comparable for detecting of the PCV
complex including the polyps [13].

6.2.2 OCT b-scans

As previously mentioned, the OCTA image set contains multiple OCT b-scans
co-registered with the OCT angiogram. It is useful to review these b-scans in concert
with the OCT angiograms. PCV is seen on OCT b-scans as elongated inverted
u-shaped pigment epithelial detachments (PED) with hyperreflective polyps adher-
ent to the underside of the PED, and an adjacent low-lying PED dubbed a ‘double
layer sign’ that represents the BVN (figure 6.3). Using the OCT b-scans alone has a
reasonably high sensitivity and specificity for detection of PCV [14, 15].

6.2.3 OCTA image set

Viewing the OCT angiogram and co-registered OCT b-scans in concert can improve
the sensitivity of detection of the PCV complex with minimal to no compromise to

Figure 6.3. The polypoidal choroidal vasculopathy complex is shown on the corresponding optical coherence
tomography b-scans as an inverted u-shaped polyp (arrowhead) adjacent to a double layer sign or low level
pigment epithelial detachment representing the branching vascular network (arrows).

6-4
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 6.4. There is a circular area of flow noted on the optical coherence tomography angiogram that could
represent a polyp (arrowhead, top). The corresponding optical coherence tomography shows an inverted
u-shaped lesion (arrowhead) adjacent to a double layer sign or low-level pigment epithelial detachment
representing the branching vascular network (arrows). The cross-hairs colocalize the circular area of flow on
the angiogram with the inverted u-shaped lesion on the b-scan increasing our suspicion that this lesion is a
polyp.

specificity (figure 6.4) [16, 17]. Oftentimes clinicians use only the OCT angiogram to
evaluate PCV, but the co-registered OCT b-scans can be an asset by providing
additional clues and signs that a PCV complex is present. The blood flow and
structural changes can be cross-referenced in order to better characterize the lesions.

6-5
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Interpreting the OCTA is limited by the presence of artifacts and inaccuracies in

the automated segmentation. These artifacts and segmentation discrepancies are
important to keep in mind as they can both falsely appear as or mask a PCV
complex. Poor signal quality due to cataracts or other media opacity can further
confuse the picture. Therefore, there is a learning curve for accurately interpreting
OCTA images and differentiating artifact from true pathology.

6.3 Discussion
In this chapter, we presented an introduction to the OCTA technology and how it
can be used to facilitate the diagnosis of PCV. As OCTA is a non-invasive and time-
efficient addition to the standard OCT machines frequently utilized to image PCV,
understanding how to use OCTA to visualize the PCV complex can be a useful
clinical skill. PCV is the most important subtype of exudative AMD, because it
predicts for anti-VEGF resistance and also can guide therapy, such as choice of anti-
VEGF therapy, and whether or not to use combination vPDT, which has been
shown to have better vision results than anti-VEGF monotherapy with one half the
number of injections for anti-VEGF monotherapy. We demonstrated that while
OCTA is limited by artifacts and the principle of slowest detectable flow, the
technology can accurately detect the PCV complex in many cases. This is
particularly true in treatment of naïve eyes and when viewing the OCTA dataset
as a whole—by using the OCT angiogram and OCT b-scans together. We described
the appearance of the polyps and the BVN on the OCTA dataset.

6.4 Conclusion
In conclusion, OCTA is a novel non-invasive technology that can be a useful tool in
the clinician’s arsenal for detecting and monitoring PCV lesions. It is important to
distinguish the PCV subtype from other forms of AMD because the treatment
algorithm may be different and it predicts anti-VEGF resistance in all ethnic
populations.

References
[1] Kokame G T, Liu K, Kokame K A, Kaneko K N and Omizo J N 2020 Clinical
characteristics of polypoidal choroidal vasculopathy and anti-vascular endothelial growth
factor treatment response in caucasians Ophthalmologica 243 178–86
[2] Pereira F B, Veloso C E, Nehemy M B and Kokame G T 2015 Characterization of
neovascular macular degeneration in Brazilian patients Ophthalmologica 234 233–42
[3] Kokame G T, deCarlo T E, Kaneko K N, Omizo J N and Lian R 2019 Anti-vascular
endothelial growth factor resistance in exudative macular degeneration and polypoidal
choroidal vasculopathy Ophthalmol. Retin. 3 744–52
[4] Lim T H, Lai T Y Y and Takahshi K et al 2020 Comparison of ranibizumab with or without
verteporfin photodynamic therapy for polypoidal choroidal vasculopathy: the EVEREST II
randomized clinical trial JAMA Ophthalmol. 138 935–42

6-6
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[5] Kokame G T, Lai J C, Wee R, Yanagihara R, Shantha J G, Ayabe J and Hirai K 2016
Prospective clinical trial of intravitreal aflibercept treatment for polyploidal choroidal vascul-
opathy with hemorrhage or exudation (EPIC study): 6 month results BMC Ophthalmol. 16 1–10
[6] de Carlo T E, Kokame G T, Shantha J G, Lai J C and Wee R et al 2018 Spectral-domain
optical coherence tomography angiography for the diagnosis and evaluation of polypoidal
choroidal vasculopathy Ophthalmologica. 239 103–9
[7] Wang M, Zhou Y and Gao S S et al 2016 Evaluating polypoidal choroidal vasculopathy
with optical coherence tomography angiography Invest. Ophthalmol. Vis. Sci. 57 526–32
[8] Takayama K, Ito Y and Kaneko H et al 2017 Comparison of indocyanine green
angiography and optical coherence tomographic angiography in polypoidal choroidal
vasculopathy Eye 31 45–52
[9] Tomiyasu T, Nozaki M, Yoshida M and Ogura Y 2016 Characteristics of polypoidal
choroidal vasculopathy evaluated by optical coherence tomography angiography Invest.
Ophthalmol. Vis. Sci. 57 324–30
[10] Kim J Y, Kwon O W and Oh H S et al 2016 Optical coherence tomography angiography in
patients with polypoidal choroidal vasculopathy Graefes. Arch. Clin. Exp. Ophthalmol. 254
1505–10
[11] Fukuyama H, Iwami H, Araki T, Ishikawa H, Ikeda N and Gomi F 2018 Indocyanine green
dye filling time for polypoidal lesions in polypoidal choroidal vasculopathy affects the
visibility of the lesions on OCT angiography Ophthalmol. Retin. 2 803–7
[12] Zhan Z, Sun L and Jin C et al 2019 Comparison between non-visualized polyps and
visualized polyps on optical coherence tomography angiography in polypoidal choroidal
vasculopathy Graefes. Arch. Clin. Exp. Ophthalmol. 257 2349–56
[13] Kim K, Yang J and Feuer W et al 2020 A comparison study of polypoidal choroidal
vasculopathy imaged with indocyanine green angiography and swept source OCT angiog-
raphy Am. J. Ophthalmol. S0002–9394 30247–6
[14] De Salvo G, Vaz-Pereira S and Keane P A et al 2014 Sensitivity and specificity of spectral-
domain optical coherence tomography in detecting idiopathic polypoidal choroidal vascul-
opathy Am. J. Ophthalmol. 158 1228–38 E1
[15] Liu R, Li J and Li Z et al 2016 Distinguishing polypoidal choroidal vasculopathy from
typical neovascular age-related macular degeneration based on spectral domain optical
coherence tomography Retina 36 778–86
[16] de Carlo T E, Kokame G T, Kaneko K N, Lian R, Lai J C and Wee R 2019 Determining the
sensitivity and specificity of detecting PCV with en-face OCT and OCTA Retina. 39 1343–52
[17] Cheung C M G, Yanagi Y and Akiba M et al 2019 Improved detection and diagnosis of
polypoidal choroidal vasculopathy using a combination of optical coherence tomography
and optical coherence tomography angiography Retina 39 1655–63

6-7
 IOP Publishing

Photo Acoustic and Optical Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
Ayman El-Baz and Jasjit S Suri

Chapter 7
Quantitative features for objective assessment of
OCT angiography
Minhaj Nur Alam, David Le, Taeyoon Son, Jennifer I Lim and Xincheng Yao

As a new quantitative imaging modality, optical coherence tomography (OCT)

angiography (OCTA) provides a non-invasive platform for detecting microvascular
abnormalities caused by different eye diseases. Since its inception, OCTA has
generated widespread interest because of its unprecedented capability to visualize
high resolution capillary structures in superficial and deep retina plexuses. In recent
years, researchers have demonstrated several quantitative imaging biomarkers for
describing retinal distortions caused by diseases such as diabetic retinopathy (DR),
diabetic macular edema (DME), age-related macular degeneration (AMD), glau-
coma, vein occlusion, sickle cell retinopathy (SCR), etc. However, it is also essential
to standardize the quantitative OCTA features for integrating them in clinical
workflow and to establish objective interpretations of clinical outcomes. In this
chapter, we summarize the state of the art of quantitative OCTA features with
technical rationale and describe potential clinical applications of quantitative OCTA
analysis. In addition, we also discuss the artificial intelligence (AI) based applica-
tions demonstrated for OCTA disease classification, feature segmentation and image
quality improvement along with their future prospect in retinal disease management.

7.1 Introduction
The retina is regularly targeted by eye diseases. As an extension of the central
nervous system (CNS), the retinal neurovascular complex can serve as a window for
assessing functionality of the brain or cardiovascular conditions. Systemic health
conditions such as diabetes can also cause retinal abnormalities such as diabetic
retinopathy (DR) and macular edema (DME). Long- term existence of diabetes can
cause hyperglycemia-induced vascular damage [1], hyperlipidemia and

doi:10.1088/978-0-7503-2060-3ch7 7-1 ª IOP Publishing Ltd 2021

 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

hypertension. Moreover, blood clots and atherosclerosis can further develop into
pathologies such as retinal vein occlusions (RVO) [2]. Therefore, in ophthalmology
practices, retinal imaging has been extensively used for screening and diagnosing
systematic diseases. Different imaging modalities include color fundus photography,
scanning laser ophthalmoscope (SLO), fluorescein angiography (FA), etc. Color
fundus photography is a widely used modality which provides valuable information
for disease detection and treatment assessment. However, the spatial resolution, field
of view and image contrast are limited to identify subtle distortions in early stages of
retinal diseases. Furthermore, many diseases manifest within the distinct retinal
layers, which cannot be observed just by visualizing the fundus of the eye. SLO [3, 4]
and adaptive optics (AO) [5–7] can provide enhanced image resolution, while FA [8, 9]
allows identification of vascular leakage and hemorrhage within the retina.
However, these imaging modalities lack the capability to differentiate individual
retinal layers and vascular plexuses. It has been established that different diseases
and their stages can target retinal vasculatures in different ways, within different
layers of the retina. Given the unprecedented capability to non-invasively differ-
entiate individual retinal layers, optical coherence tomography (OCT) [10] has been
widely employed for depth-resolved investigation of morphological abnormalities
caused by different retinal diseases [11–13]. OCT angiography (OCTA) is a recent
OCT imaging modality that can provide a non-invasive method to visualize blood
flow information in individual capillary plexus layers in the outer and inner retina
[14, 15] with high resolution. Since its first commercial launch in 2014, OCTA has
rapidly demonstrated its excellence as a quantitative imaging technique in clinical
management of diabetic retinopathy (DR) [16, 17], glaucoma [18, 19], sickle cell
retinopathy (SCR) [20], age-related macular degeneration (AMD) [21] and other eye
diseases. Standardized quantitative OCTA analysis is essential for objective inter-
pretation of clinical outcomes and designing treatment strategies. Researchers have
developed multiple OCTA features in recent years for quantitative analysis of
vascular distortions caused by a myriad of eye conditions and systematic diseases
[22]. In this chapter, we discuss, summarize and provide technical rationales of these
quantitative OCTA features for a better understanding of current trend and state of
the art. Additionally, current status and future prospects of using OCTA features for
artificial intelligence (AI) based objective detection and classification of eye diseases
or specific features will be discussed in brief.

7.2 Quantitative OCTA features

We will briefly introduce and provide brief technical rationale of different OCTA
features developed in literature, including blood vessel density (BVD), blood
vessel caliber (BVC), vessel perimeter index (VPI), blood vessel tortuosity (BVT),
fovea avascular zone features (area-FAZ-A, contour irregularity-FAZ-CI), vessel
complexity features, branchpoint features (BPF), differential artery and vein
(A-V) features, and features based retinal flow analysis and choroidal vasculature
quantification.

7-2
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

7.2.1 Blood vessel density

BVD is one of the most commonly analyzed OCTA features, and has been
demonstrated to be associated with abnormalities associated with diseases such as
DR [23, 24], SCR [25, 26], AMD [27, 28], glaucoma [29, 30] and vein occlusion (VO)
[31, 32] which induce vessel abnormalities such as ischemia and drop-out areas in retinal
vasculatures. Most of these diseases initially manifest at the capillary level, which can be
detected in OCTA. Similar features that have been demonstrated in literature with
different terminology include vessel density (VD) [33], percent area of nonperfusion
(PAN) [34], capillary density (CD) [35], vessel area density (VAD) [36], etc. All together
these density features reflect the ratio of the image area that is occupied by blood vessels
(figure 7.1(b)). BVD can be quantified with the following equation [37]:
n
∑ A(x , y )
x = 1, y = 1
BVD = n (7.1)
∑ I (x , y )
x = 1, y = 1

where A(x,y) is all the pixels occupied by the blood vessels, and I(x,y) is all the pixels
in the OCTA image. Similar metrics can be measured for vessel skeletons density as
well, where A can be replaced with S (skeleton pixels).

7.2.2 Blood vessel caliber

BVC is a metric that is used to quantify the dilation or shrinkage of the blood
vessels. Similar studies have also referred to BVC as vessel diameter, vessel width or
vessel diameter index (VDI) [38]. BVC distortions are prevalent in patients with
hypertension, DR [39] and SCR [37, 40]. BVC is defined as the ratio of the vascular
area to the total vascular length [37]:
n
∑ A(x , y )
x = 1, y = 1
BVC = n (7.2)
∑ S (x , y )
x = 1, y = 1

Figure 7.1. OCTA Feature extraction. (a) A representative 6 mm × 6 mm OCTA superficial scan from a DR
patient. (b) Extracted vessel map showing segmented fovea in orange. (c) Skeletonized vessel map. (d) Vessel
perimeter map. (e) Fractal contour map.

7-3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

where A(x,y) is the pixels in the segmented vessel map (figure 7.1(b)) and S(x,y) is the
n
pixels in the skeletonized vessel map (figure 7.1(c)). So, ∑ A(x , y ) represents the
x = 1, y = 1
n
total vascular area, and ∑ S (x , y ) represents the total vessel length.
x = 1, y = 1

7.2.3 Blood vessel tortuosity

BVT [37, 40] is a measure of the geometric tortuosity change in blood vessels. It can
be correlated to the degree of individual vessel distortion. Normally, apart from the
regular patterning of the blood vessel network, for efficient blood/oxygen transport,
the blood vessels have a relatively smooth structure. In diseased conditions [37, 39–42],
the epithelial tissue surrounding the blood vessel walls are affected or in some cases,
distorted shape of blood cells cause occlusion and distortions in the structure of the
blood vessels, affecting the blood/oxygen transportation efficiency of blood vessel
network. This structural distortion can be correlated to the vessel tortuosity change,
which can be quantified by taking the ratio of the geodesic and Euclidean distance
[37] between two vessel endpoints:

1
BVT =
n
n
⎛ ⎞ (7.3)
∑⎜ Geodesic distance between two endpoints of a vessel branch i ⎟
i = 1 ⎝ Euclidean distance between two endpoints of a vessel branch i ⎠

where i is equal to the ith branch and n is equal to the total number of branches;
Euclidian distance here represent the straight-line distance between two vessel end
points, whereas Geodesic distance represents the total curve length between two
vessel end points (as shown in figure 7.1(c)).

7.2.4 Vessel perimeter index

VPI [43] is a metric that measures the ratio between overall vessel boundary
(figure 7.1(d)) and total blood vessel area in segmented vessel maps (figure 7.1(b)).
VPI has been demonstrated to reflect vessel dropout and early ischemia [39, 44], in
DR [44] and SCR [40]. It can be measured using the following equation [37]:
n
∑ P (x , y )
x = 1, y = 1
VPI = n (7.4)
∑ I (x , y )
x = 1, y = 1

where P(x,y) is the vessel perimeter pixels and I(x,y) is all the vessel pixels.

7-4
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

7.2.5 Foveal avascular zone features

The fovea is one of the most affected areas in the retina by retinal diseases such as
DR [45], VO [46], and SCR [40] due to foveal ischemia and parafoveal vessel drop-
outs. FAZ-A is a sensitive feature that can quantify the shape change of the fovea
and has been demonstrated to be able to differentiate the severities of different non-
proliferative diabetic retinopathy (NPDR) stages [39]. In general, FAZ-A is
determined by first segmenting the FAZ (blue region in figure 7.1(c)) and then
calculating the total segmented FAZ area with the following equation [37]:
n
FAZ − A = [Area of single pixel(in μm2) × ∑ A(x , y )] (7.5)
x = 1, y = 1

where A(x,y) is the segmented FAZ pixels.

FAZ-CI, is another foveal feature, that measures the structural irregularity of the
fovea [40]. Recent studies has been also termed it as FAZ-circularity [40] or FAZ
acircularity index (FAZ-AI) [47] and demonstrated FAZ-CI to be sensitive to
vascular contortions caused by DR [47], AMD [48], SCR [37], and glaucoma [49].
The FAZ-CI can be measured as the ratio of the foveal perimeter and the perimeter
of a reference circle that has an area identical to the fovea [37]:
n1
∑ O (x , y )
x = 1, y = 1
FAZ − CI = n2
(7.6)
∑ R(x , y )
x = 1, y = 1

where O(x,y) is the foveal perimeter pixel (green demarcation—figure 7.1(c)), R(x,y)
is the reference circle pixel.

7.2.6 Vessel complexity features

Vessel complexity features have been shown to be sensitive in DR detection, in
particularly for monitoring the progression of DR from the non-proliferative to the
proliferative stage [50]. Vessel complexity index (VCI) is one of the complexity
features and can be quantified from the perimeter map and vessel map [25]:
(Pixels enclosed by perimeter map)2
VCI = (7.7)
4π × (Pixels enclosed by vessel map)
VCI uniquely enables the localization of complex vascular morphologies within the
retina.
Similarly, fractal dimension (FD) has also been used to quantify the complexity of
the vasculature (figure 7.1(e)). FD is generally calculated using the box-counting
method, which establishes a relationship between the number of boxes (of a
certain size/scale) that can enclose the pattern within an image (figure 7.1(e)) [51].

7-5
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

This box-count can be iteratively measured for different scales. The box-counting
method is equated as [51]:
logNr
FD = (7.8)
logr −1
where Nr is the number of boxes which enclose the image pattern by the scale, r.
Another complimentary parameter to FD is Lacunarity (LAC) which is a measure
of rotational inhomogeneity between vessel structures [52], specifically LAC
provides information regarding the size of the gaps and their distribution around
objects of interest in binary images.

7.2.7 Branchpoint analysis

Decreased efficiency in blood/oxygen transport in the retina can affect bifurcation
in vascular structures [53]. Quantitative OCTA branchpoint analysis with BPFs,
including both angle- and width-based features (figure 7.2), has been recently
explored for objective detection of DR [54]. Angle parameters including the vessel
branching angle (VBA) and the child branching angles (CBA1 and CBA2),
quantify in vessel bifurcation changes [54]. To measure the branching angles,
first the branchpoint and branch endpoints must be determined (green dots
and red dots in figures 7.2(b) and (c), respectively). Then the branchpoint angles
are measured. Analogously, two width-based parameters were also quantified,
namely the vessel branching coefficient (VBC) and child width ratios (CWRs).

Figure 7.2. (a) Sample branchpoint; (b) Branchpoint showed in a vessel skeleton. Here the green dot represents
a branchpoint and the red dots represent three end points, the blue pixels represent the parent and children
vessels of interest, and the yellow circle shows the dilated area. (c) Identified branchpoint (green) and endpoints
(red), the yellow square is the window area. (d) Branch angle measurement. Modified from [54].

7-6
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Width-based features examine the structural change of the blood vessels as a

result of bifurcation [54]:
d12 + d 22
VBC = (7.9)
d 02

di
CWRi = (7.10)
d0
where d 0 is parent vessel width, d1 and d2 are the child vessel widths, and i represents
the ith child vessel. As a result of ischemia and neovascularization, changes in
branching angles and widths could be correlated to vascular remodeling.

7.2.8 Flow analysis

Several parameters have been developed to quantify the blood flow in retinal tissues,
namely the flow index (FI) [55], flow void (FV) [56], adjusted flow index (AFI) [34],
vascular connectivity [57], and total retinal blood flow (TRBF). FI is defined as the
average OCTA decorrelation value within a region in the retina [58]. The FI can be
measured using the following equation,
n
∑ A(x , y )
(7.11)
x = 1, y = 1
FI =
N
where A(x,y) is the pixel decorrelation value, and N is the total number of OCTA
pixels.
The FV is measured as the percentage of the non-flow area over the total scanned
region [56]
AreaFlowvoid
FV = × 100% (7.12)
Area whole
TRBF is a parameter demonstrated using Doppler OCT and can be used to
estimate the flow in all the vessel segments [59]:

TRBF = − ∬ vz(x , y )dxdy (7.13)

xy − plane

where the flow in a blood vessel is computed by integrating the axial flow velocity,
vz (x , y ) measured from the different pixels in the OCTA image over the surface
normal to vz (x , y ) [59]. This method is typically applied in OCTA systems with high
acquisition speeds (e.g. 100 kHz).
Another OCTA parameter related to flow is named variable interscan time
analysis (VISTA) [60, 61], which can be performed to assess the changes in
choriocapillaris for differentiating varying degrees of flow impairment. In different

7-7
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

analysis, VISTA has been demonstrated to quantify changes in AMD, geographic

atrophy (GA) and NPDR/PDR eyes.

7.2.9 Choroidal vasculature

Aside from the retinal vasculature, OCTA features have also been demonstrated for
analyzing choroidal vasculature. For example, qualitative observations of CNV
using OCTA have revealed distinct morphological distortions such as medusa,
seafan, tangled, and dead-tree type patterns within the choroidal vasculature [28].
CNV area is a standardized parameter and can be measured as follows [62]:
CNV area(mm2) = CNV area(pixel) × (Field of view in mm/304 pixel)2 (7.14)
The first step for CNV analysis is a reliable segmentation (figure 7.3). In literature,
the segmentation has typically been performed using manual or semi-manual
methods [14]. Recently, projection-resolved (PR) OCTA techniques have been
explored to determine the CNV area and skeleton automatically [57]. The binarized
CNV vessel or skeleton map can be used for FD analysis [28], and measurement of
BVD or LAC. Other CNV features include CNV maturity, location, presence of
core vessels and loops [63], etc.

7.2.10 Differential artery–vein (A–V) analysis

Differential A–V analysis in OCTA compares the changes due to diseases that affect
artery and veins differently. There have been only few studies that demonstrated

Figure 7.3. Illustration of different image processing steps for generating the CNV area and its skeleton from
an original outer retinal en face OCTA choroidal image using saliency model. Reprinted from [57].

7-8
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

A–V classification in OCTA. The first study employed color fundus image guided A–V
classification in OCTA (figure 7.4). Differential A–V analysis revealed improved
sensitivity to identify DR [42] and SCR [37, 40] associated changes. One limitation of
these methods is the use of two imaging modalities and image registration. This
limitation is addressed in following studies by replacing color fundus image with
en face OCT images, which retain some intensity profile information to identify A–V
near the optic disks (figure 7.5) [64]. Using en face OCT for A–V classification in
OCTA could improve the efficiency for clinical deployment of differential A–V
analysis. Since OCT and OCTA images are intrinsically reconstructed from the
same raw spectrogram, it essentially removes any requirement for image registration,
and enables single instrument deplyment of OCTA classification. Apart from this,
another method for A–V classification in OCTA was demonstrated using near infrared
oximetry from OCT images [65] (figure 7.6). Son et al [66] reported an automated
method for OCTA A–V classification by employing near-infrared OCT oximetry to
guide the A–V classification in macular OCTA scans. This study developed a custom

Figure 7.4. Fundus guided A–V classification in OCTA [42]. (a) Fundus image. (b) Corresponding OCTA
image; (c) A–V map of an OCTA image overlaid on the fundus image. Modified from [41].

Figure 7.5. En face OCT guided A–V classification in OCTA [64]. (a) An en face OCT vessel map in which
source nodes have been classified into artery–vein (b) en face OCT A–V map. (c) OCTA image, (d) OCTA
binary map, (e) en face OCT A–V map overlaid on the OCTA binary map, (f) final OCTA A–V map
(Modified from [64]).

7-9
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 7.6. Oxymetry guided A–V classification. (a1) OCT B-scan. (a2) En face OCT. (a3) Macular OCTA
scan. (b) En face OCT large field of view image covering optic disc and foveal region. (c1) En face OCT image
of optic disc. (c2) Sampling locations of tissue and vessel area. (c3) Intensity profile of the vessels and tissue.
(d1–d2) ODRs from blood vessels in (c1) at 765 nm and 855 nm wavelength. (e1–e2) ODRs from each subject
with 765/805 nm and 805/855 nm analysis. Solid black lines show averaged ODR of all arteries’ and veins’
ODRs for each subject. (f) Averaged ODR difference between artery and vein. (Reprinted from [66].)

OCT/OCTA device which used an oxygen sensitive 765 nm wavelength that provided
2.8× higher oxygen absorption coefficient between artery and vein (figure 7.6).

7.3 Machine learning techniques in objective OCTA analysis

Machine learning techniques have garnered huge attention in the field of ophthal-
mology in recent years. This includes supervised techniques that rely on quantitative
biomarkers or features that correlate to pathophysiological changes in the retina,
and more recently unsupervised ‘deep’ machine learning methods uses representa-
tion-learning to identify intricate structures in high-dimensional data. However,
the use of machine learning has been primarily limited to fundus images because
fundus data is widely available due to its regular use in clinical ophthalmic practices.

7-10
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Some studies have applied machine learning techniques in OCT and achieved
excellent results for layer segmentation [67–69]. Only recently have we seen some
machine learning applications in OCTA [37, 39]. The limitation of database size is a
major issue for OCTA. However, because the OCTA quantitative features have
shown remarkable results in identifying disease onset and progression of stages, a
supervised machine learning technique can be implemented for preliminary studies
using OCTA. In recent studies, Alam et al used a support vector machine (SVM)
classifier for objective classification of SCR and DR. In the first study, six OCTA
features, i.e., BVD, BVC, BVT, VPI, FAZ-A, and FAZ-CI, were used to train SVM
classifiers for identifying control, mild and severe SCR subjects. The SVM was able
to identify control versus disease and mild versus severe SCR with 100% and 97%
accuracies, respectively. In a subsequent study, they adapted a similar approach for
classification of NPDR and observed 94.41% accuracy for control versus disease (i.e.
NPDR) and 92.96% accuracy for control versus mild NPDR classification. The
control versus mild NPDR was significantly important because it is crucial to
identify the onset of NPDR in patients with DM. In a similar study, Sandhu et al
[70] used three OCTA features, i.e., BVD, BVC and DAZ-A, for automated
classification of NPDR with 94.3% accuracy.
In both studies, the best results were obtained using all OCTA features as a
combined set to train SVM classifiers. However, the accuracy value for a combined
set was only compared with individual features. A more optimal solution would be
to choose the best combination of OCTA feature-sets for classification tasks. For
example, Ashraf et al conducted a study with multiple OCTA features to find a
statistically significant combination of features to distinguish DR subjects from
controls. Their study employed a regression model to identify superficial plexus
FAZ-A, deep plexus VD and FAZ-CI as the best combination for objective
classification. Similarly, in a recent study, Alam et al proposed a multi-task AI
classifier (sample workflow in figure 7.7), which used a logistic regression based
backward elimination technique to identify an optimized feature set for multiple
tasks, such as control versus disease, inter-disease and disease staging classification.
For a proof of method study, only DR and SCR OCTAs from a retinal clinic were
used. The classifier identified perifoveal BVD (superficial plexus), FAZ-A (super-
ficial plexus) and FAZ-CI (deep plexus) for control versus disease classification;

Figure 7.7. Supervised machine learning for objective classification of OCTA.

7-11
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

BVT (superficial plexus), BVD (superficial plexus), FAZ-A (superficial plexus), and
FAZ-CI (deep plexus) for DR versus SCR classification; BVD (superficial plexus)
and FAZ-A (superficial plexus) for NPDR staging; and BVT (superficial plexus),
BVD (superficial plexus), and FAZ-CI (superficial plexus) for SCR staging.
This study demonstrated the importance of identifying the most sensitive features
for different retinopathies. Furthermore, by choosing the optimized features from
both superficial and deep plexuses, the AI-based classifier ensured representation
of layer specific abnormalities due to retinopathies. Although machine learning
based OCTA studies showed excellent preliminary results, they were not validated
with independent clinical data. Therefore, the generalized performance of OCTA
classification on a population of different races and ethnicities could not be
analyzed. Further studies with multi-center OCTA data and multi-device OCTA
images are required for clinical validation of AI-based OCTA classification.
The use of unsupervised and deep machine learning on OCTA has recently been
demonstrated, however, it remains not fully explored due to limited availability of
large datasets. Initial studies have used convolutional neural networks (CNNs) for
DR detection [71]. Le and Alam et al demonstrated a transfer learning based
approach for identifying control and three NPDR stages [72]. Aside from disease
diagnosis, few studies have implemented deep machine learning for image process-
ing-based applications in OCTA. For example, Prentasic et al [73] demonstrated a
convolutional neural network (CNN) based technique to segment microvasculature
in a foveal area. They use 80 OCTA images and a cross validation technique to test
their CNN on identifying vessel and non-vessel pixels in a parafovea OCTA scan.
Compared to manual segmentation results, this study showed 83% accuracy for
vessel segmentation. When the result was compared with inter- and intra-rater
accuracies, it was observed that the CNN was equivalent to a second manual rater.
A similar application was adapted by Guo et al [74] to segment retinal non-prefusion
areas in ultra-wide-field angiograms. In a follow-up study, they further demon-
strated a CNN that was able to identify non-perfusion from signal reduction
artifacts [75]. In a separate study, Kasaragod et al [76] used machine learning based
techniques for segmentation of the optical nerve head (ONH). However, they used
Jones matrix OCT (JM-OCT), which provides OCT, OCTA, birefringence tomog-
raphy and attenuation coefficient tomography at the same time. The semi-automatic
unsupervised classification framework was specially designed for ONH tissue
segmentation. Deep learning has been also widely used for image processing
applications in OCT images, as previously mentioned. OCT layer segmentation is
a major application. Although not directly a machine learning application to
OCTA, OCT retinal layer segmentation can be correlated to OCTA superficial
and deep plexus segmentation, which in turn provides better image quality from
OCTA. Fang et al demonstrated a deep learning based strategy to identify nine
retinal layers in OCT B-scans [68]. This information could be transferred to
OCTA B-scan volumes for better segmentation of inner and outer plexiform
layers, which include retinal information. The clinical OCTA devices can
integrate robust segmentation for removing any shadow artifacts in OCTA
en face projection images.

7-12
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

7.4 Discussion
Compared to imaging modalities like fundus photography, SLO, and FA, OCTA
provides a label-free solution for high resolution visualization and quantification of
retinal vasculatures. With the capability to image depth-resolved retinal vasculatures
at capillary level resolution, OCTA is being rapidly adopted for the clinical
management of different eye diseases. We have summarized the use of quantitative
OCTA features in table 7.1, organizing literature which utilized commercial devices
and software. Different quantitative features have been demonstrated and widely
explored to foster the standardization of objective OCTA interpretation.
It has been observed that pathological mechanisms of different eye diseases can
affect the sensitivity of the OCTA features. For example, SCR produces sickle
shaped blood cells, that induces tortuous and dilated vessels, which can be quantified
using BVT and BVC, respectively. From OCTA analysis of SCR patients, BVT has
been demonstrated as the most sensitive feature for SCR classification [77].
Analogously, DR patients frequently have associated hypertension which can cause
arterial narrowing [42], which can be measured using BVC. BVD analysis in DR
patients can also assess capillary level ischemia in DR patients [39]. BVD is also
sensitive enough to evaluate central/branch VOs [32, 46, 78–80]. Recent studies also
showed that vascular distortions could manifest in different localized regions due to
different diseases. For example, localized VCI measurements have been demon-
strated to be successful at identifying DR progression to PDR [50]; BVD measure-
ments also revealed that the perifoveal region is the most sensitive region for
classifying NPDR stages [39, 81]. On the other hand, for SCR, the parafoveal
temporal retina is the most sensitive local region for SCR staging [40]. This creates
the possibility for a multitude of interesting investigations of OCTA features to
customize and tailor features based on disease pathology.
Differential A–V analysis is also one of the new developments in quantitative
OCTA analysis that has been demonstrated to improve OCTA feature sensitivity for
retinopathy staging [41, 42]. Pathological deformations in the artery–vein morphol-
ogy can be affected in distinct trends [82]. DR may induce increased venous beading,
venous dilation, arterial narrowing and arterial tortuosity [83]. The sensitivity of the
OCTA features may be compromised if an average global value of A–V distortions
is compared between healthy and diseased eye. Two recent studies revealed two
differential A–V features, i.e. A–V ratio of BVC (AVR–BVC) and BVT (AVR–
BVT) to improve the performance for DR and SCR staging [41, 42].
For AMD assessment, predominantly quantitative analyses of choroidal
vasculature have been explored [60, 84]. Early stage or dry AMD is normally
characterized by decreased choroidal blood flow and drusen. More advanced dry
AMD is characterized by geographic atrophy. In the case of late stage or wet
AMD, CNVs are the most common biomarkers [60, 85]. Jia et al first presented an
OCTA study to detect and classify CNV types (I and II) [17], reporting decreased
choroidal flow near the CNV in all cases. In following studies [57], for quantifi-
cation of CNV area and connectivity, researchers analyzed artifact removal
algorithms, and projection resolved OCTA (PR-OCTA). Miller et al compared

7-13
 Table 7.1. Summary of quantitative OCTA features.

Clinical OCTA with

Features Applications built-in software Clinical OCTA with custom software Custom OCTA

BVD To quantify vessel density and [27, 33, 48, 52, 87–98] [23, 28–31, 34, 36–40, 42, 43, 46, 70, 81, 93, [134, 135]
identify ischemic regions 95, 99–133]
BVC To quantify blood vessel width, [136] [36–42, 44, 70, 106, 108, 109, 111, 116, 117, [134, 135, 139]
shrinkage or dilation 120, 124, 131, 137, 138]
BVT To quantify blood vessel tortuosity [140] [24, 31, 37, 39–42, 119, 123, 138, 141–144] [135]
because of change in morphology
— —
VPI To quantify changes in the blood [36, 37, 39, 40, 43, 116, 117, 124, 131, 133]
vessel perimeter
—
FAZ-A To quantify area changes in foveal [24–27, 32, 33, 45, 48, 52, [24, 31, 34, 37, 39, 40, 44–47, 70, 81, 93, 95,
78, 87–94, 96, 98, 136, 101, 110–115, 119, 121–123, 125, 127,
140, 145–157] 129, 131, 140, 152, 158–169]
—
FAZ-CI To quantify complexity changes in [48, 87, 88, 91–93, 156, [47, 81, 110, 111, 113, 122, 131, 152, 159,

7-14
foveal contour 164] 162, 163, 165]
— —
VCI To quantify complexity changes in [28, 30, 38, 51, 81, 99–109, 111, 124, 126,
retinal vasculature 170–175]
— —
BPA To quantify changes in vascular [54]
branching geometry bifurcation
—
A-V analysis To achieve differential artery and [41, 42, 137] [65]
vein analysis in DR, and SCR, etc
Flow analysis To quantify blood flow changes [33, 145] [34, 55–57, 128, 130, 158, 176, 177] [17, 60, 178]
—
CNV analysis To quantify neovascularization [28, 57, 62, 63, 84, 86, 102, 103, 138, 170, [17, 60, 86]
Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

morphology change 172, 177, 179]

BVD: blood vessel density; BVC: blood vessel caliber; BVT: blood vessel tortuosity; VPI: vessel perimeter index; FAZ: foveal avascular zone; CI: contour irregularity;
FD: fractal dimension; VCI: vessel complexity index; LAC: lacunarity; BPA: branchpoint analysis; A–V: artery–vein; AVR: artery–vein ratio; CNV: choroidal
neovasculature..
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

CNV using spectral domain (SD) and swept source (SS)-OCTA [86], and reported
statistically significant differences between lesion areas measured by SS and SD
OCTA. SS-OCTA showed larger lesion area in both 3 × 3 mm2 and 6 × 6 mm2,
and larger differences were reported in the 6 × 6 mm2 OCTA scans. Another
parameter studied for AMD is FD; Al-Sheikh et al studied CNV lesions in pre-
and post-treatment AMD and reported a lower FD value in the inner part of the
lesion post treatment [28].
Quantitative OCTA has opened a unique opportunity for computer-aided
detection and AI classification in different eye diseases, as described in section 7.3
in detail. Machine learning techniques have been also explored to segment micro-
vasculature [73], non-prefusion areas [74], optical nerve head (ONH) [76]. Although,
the limited size of currently available OCTA database remains a major challenging
for deep learning-based studies with OCTA, with the increasing interest in AI
applications in ophthalmology, there will be a widespread incentive to expand the
available OCTA datasets and multi-center collaborations. Therefore, we anticipate
that deep learning-based classification (already well established for fundus image
application) will play an important role to enable future investigations of automated
OCTA detection and classification of eye conditions. Concurrently, transfer learning
based technology [180] might also find valuable applications in the near future to
foster deep learning based OCTA studies.

References
[1] Wang W and Lo A C 2018 Diabetic retinopathy: pathophysiology and treatments Int. J.
Mol. Sci. 19 1816
[2] Karia N 2010 Retinal vein occlusion: pathophysiology and treatment options Clin.
Ophthalmol. 4 809
[3] Sharp P F and Manivannan A 1997 The scanning laser ophthalmoscope Phys. Med. Biol.
42 951–66
[4] Kobayashi K and Asakura T 1995 Imaging techniques and applications of the scanning
laser ophthalmoscope Opt. Eng. 34 717–26
[5] Liang J Z, Williams D R and Miller D T 1997 High resolution imaging of the living human
retina with adaptive optics Investig. Ophthalmol. Vis. Sci. 38 55
[6] Burns S A, Elsner A E, Sapoznik K A, Warner R L and Gast T J 2019 Adaptive optics
imaging of the human retina Prog. Retin. Eye Res. 68 1–30
[7] Paques M, Meimon S, Rossant F, Rosenbaum D, Mrejen S, Sennlaub F and Grieve K 2018
Adaptive optics ophthalmoscopy: application to age-related macular degeneration and
vascular diseases Prog. Retin. Eye Res. 66 1–16
[8] Banda H K, Shah G K and Blinder K J 2019 Applications of fundus autofluorescence and
widefield angiography in clinical practice Can. J. Ophthalmol. 54 11–9
[9] Mehta S 2006 Fundus fluorescein angiography of choroidal tubercles: case reports and
review of literature Indian J. Ophthalmol. 54 273–5
[10] Huang D et al 1991 Optical coherence tomography Science 254 1178–81
[11] Mailankody P, Lenka A and Pal P K 2019 The role of optical coherence tomography in
parkinsonism: a critical review J. Neurol. Sci. 403 67–74

7-15
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[12] Ruia S, Saxena S, Gemmy Cheung C M, Gilhotra J S and Lai T Y 2016 Spectral domain
optical coherence tomography features and classification systems for diabetic macular
edema: a review Asia Pac. J. Ophthalmol. (Phila.) 5 360–7
[13] Sohrab M A and Fawzi A A 2013 Review of en-face choroidal imaging using spectral-
domain optical coherence tomography Med. Hypothesis Discov. Innov. Ophthalmol. 2 69–73
[14] Gao S S, Jia Y, Zhang M, Su J P, Liu G, Hwang T S, Bailey S T and Huang D 2016 Optical
coherence tomography angiography Invest. Ophthalmol. Vis. Sci. 57 OCT27–36
[15] Zhang A, Zhang Q, Chen C L and Wang R K 2015 Methods and algorithms for
optical coherence tomography-based angiography: a review and comparison J. Biomed.
Opt. 20 100901
[16] Bandello F, Corbelli E, Carnevali A, Pierro L and Querques G 2016 Optical coherence
tomography angiography of diabetic retinopathy Dev. Ophthalmol. 56 107–12
[17] Jia Y, Bailey S T, Wilson D J, Tan O, Klein M L, Flaxel C J, Potsaid B, Liu J J, Lu C D
and Kraus M F 2014 Quantitative optical coherence tomography angiography of choroidal
neovascularization in age-related macular degeneration Ophthalmology 121 1435–44
[18] Dastiridou A and Chopra V 2018 Potential applications of optical coherence tomography
angiography in glaucoma Curr. Opin. Ophthalmol. 29 226–33
[19] Bojikian K D, Chen P P and Wen J C 2019 Optical coherence tomography angiography in
glaucoma Curr. Opin. Ophthalmol. 30 110–6
[20] Leitao Guerra R L, Leitao Guerra C L, Bastos M G, de Oliveira A H P and Salles C 2019
Sickle cell retinopathy: what we now understand using optical coherence tomography
angiography. A systematic review Blood Rev. 35 32–42
[21] Faatz H, Farecki M L, Rothaus K, Gunnemann F, Gutfleisch M, Lommatzsch A and
Pauleikhoff D 2019 Optical coherence tomography angiography of types 1 and 2 choroidal
neovascularization in age-related macular degeneration during anti-VEGF therapy: eval-
uation of a new quantitative method Eye 33 1466–71
[22] Yao X, Alam M N, Le D and Toslak D 2020 Quantitative optical coherence tomography
angiography: a review Exp. Biol. Med. (Maywood) 245 301–12
[23] Pedinielli A, Bonnin S, Sanharawi M E, Mane V, Erginay A, Couturier A and Tadayoni R
2017 Three different optical coherence tomography angiography measurement methods for
assessing capillary density changes in diabetic retinopathy Ophthalmic Surg. Lasers Imaging
Retina 48 378–84
[24] Mastropasqua R, Toto L, Mastropasqua A, Aloia R, De Nicola C, Mattei P A, Di Marzio
G, Di Nicola M and Di Antonio L 2017 Foveal avascular zone area and parafoveal vessel
density measurements in different stages of diabetic retinopathy by optical coherence
tomography angiography Int. J. Ophthalmol. 10 1545
[25] Falavarjani K G, Scott A W, Wang K, Han I C, Chen X, Klufas M, Hubschman J-P,
Schwartz S D, Sadda S R and Sarraf D 2016 Correlation of multimodal imaging in sickle
cell retinopathy Retina 36 S111–7
[26] Minvielle W, Caillaux V, Cohen S Y, Chasset F, Zambrowski O, Miere A and Souied E H
2016 Macular microangiopathy in sickle cell disease using optical coherence tomography
angiography Am. J. Ophthalmol. 164 137–44 e131
[27] Jhaj G, Glazman S, Shrier E M and Bodis-Wollner I 2017 Non-exudative age-related
macular degeneration foveal avascular zone area, foveal vessel density, and ganglion cell
complex thickness Invest. Ophth. Vis. Sci. 58 36

7-16
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[28] Al-Sheikh M, Iafe N A, Phasukkijwatana N, Sadda S R and Sarraf D 2018 Biomarkers of

neovascular activity in age-related macular degeneration using optical coherence tomog-
raphy angiography Retina 38 220–30
[29] Liu L, Jia Y, Takusagawa H L, Pechauer A D, Edmunds B, Lombardi L, Davis E,
Morrison J C and Huang D 2015 Optical coherence tomography angiography of the
peripapillary retina in glaucoma JAMA Ophthalmol. 133 1045–52
[30] Richter G M, Chang R, Situ B, Chu Z, Burkemper B, Reznik A, Bedrood S, Kashani A H,
Varma R and Wang R K 2018 Diagnostic performance of macular versus peripapillary
vessel parameters by optical coherence tomography angiography for glaucoma Transl. Vis.
Sci. Technol. 7 21
[31] Adhi M et al 2016 Retinal capillary network and foveal avascular zone in eyes with vein
occlusion and fellow eyes analyzed with optical coherence tomography angiography Invest.
Ophthalmol. Vis. Sci. 57 OCT486–494
[32] Winegarner A et al 2018 Changes in retinal microvasculature and visual acuity after
antivascular endothelial growth factor therapy in retinal vein occlusion Invest. Ophthalmol.
Vis. Sci. 59 2708–16
[33] Chao S-C, Yang S-J, Chen H-C, Sun C-C, Liu C-H and Lee C-Y 2019 Early macular
angiography among patients with glaucoma, ocular hypertension, and normal subjects J.
Ophthalmol. 2019 7419470
[34] Nesper P L, Roberts P K, Onishi A C, Chai H, Liu L, Jampol L M and Fawzi A A 2017
Quantifying microvascular abnormalities with increasing severity of diabetic retinopathy
using optical coherence tomography angiography Invest. Ophth. Vis. Sci. 58 BIO307–15
[35] Camino A, Zhang M, Liu L, Wang J, Jia Y and Huang D 2018 Enhanced quantification of
retinal perfusion by improved discrimination of blood flow from bulk motion signal in
OCTA Transl. Vis. Sci. Technol. 7 20
[36] Lin T-C, Gogte P, Palejwala N, Shahidzadeh A, Itty S, Humayun M S, Moshfeghi A A,
Ameri H, Chu Z and Wang R K 2017 Quantitative spectral-domain optical coherence
tomography angiography (OCTA) of diabetic retinopathy (DR) severity Invest. Ophth. Vis.
Sci. 58 1653
[37] Alam M, Thapa D, Lim J I, Cao D and Yao X 2017 Computer-aided classification of sickle
cell retinopathy using quantitative features in optical coherence tomography angiography
Biomed. Opt. Express 8 4206–16
[38] Akagi T, Uji A, Huang A S, Weinreb R N, Yamada T, Miyata M, Kameda T, Ikeda H O and
Tsujikawa A 2018 Conjunctival and intrascleral vasculatures assessed using anterior segment
optical coherence tomography angiography in normal eyes Am. J. Ophthalmol. 196 1–9
[39] Alam M, Zhang Y, Lim J I, Chan R V, Yang M and Yao X 2019 Quantitative optical
coherence tomography angiography features for objective classification and staging of
diabetic retinopathy Retina 40 322–32
[40] Alam M, Thapa D, Lim J I, Cao D and Yao X 2017 Quantitative characteristics of sickle
cell retinopathy in optical coherence tomography angiography Biomed. Opt. Express 8
1741–53
[41] Alam M, Lim J I, Toslak D and Yao X 2019 Differential artery–vein analysis improves the
performance of OCTA staging of sickle cell retinopathy Transl. Vis. Sci. Technol. 8 3
[42] Alam M, Toslak D, Lim J I and Yao X 2018 Color fundus image guided artery-vein
differentiation in optical coherence tomography angiography Investig. Ophthalmol. Vis. Sci.
59 4953–62

7-17
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[43] Ye H, Zheng C, Lan X, Zhao L, Qiao T, Li X and Zhang Y 2019 Evaluation of retinal
vasculature before and after treatment of children with obstructive sleep apnea-hypopnea
syndrome by optical coherence tomography angiography Graefe’s Arch. Clin. Exp.
Ophthalmol. 257 543–8
[44] Chu Z, Lin J, Gao C, Xin C, Zhang Q, Chen C-L, Roisman L, Gregori G, Rosenfeld P J
and Wang R K 2016 Quantitative Assessment of the Retinal Microvasculature Using Optical
Coherence Tomography Angiography (Bellingham, WA: SPIE)vol 21 1–13 13
[45] Durbin M K, An L, Shemonski N D, Soares M, Santos T, Lopes M, Neves C and Cunha-
Vaz J 2017 Quantification of retinal microvascular density in optical coherence tomo-
graphic angiography images in diabetic retinopathy JAMA Ophthalmol. 135 370–6
[46] Falavarjani K G, Iafe N A, Hubschman J-P, Tsui I, Sadda S R and Sarraf D 2017 Optical
coherence tomography angiography analysis of the foveal avascular zone and macular
vessel density after anti-VEGF therapy in eyes with diabetic macular edema and retinal vein
occlusion Invest. Ophth. Vis. Sci. 58 30–4
[47] Krawitz B D, Mo S, Geyman L S, Agemy S A, Scripsema N K, Garcia P M, Chui T Y and
Rosen R B 2017 Acircularity index and axis ratio of the foveal avascular zone in diabetic
eyes and healthy controls measured by optical coherence tomography angiography Vis.
Res. 139 177–86
[48] Li Z, Alzogool M, Xiao J, Zhang S, Zeng P and Lan Y 2018 Optical coherence tomography
angiography findings of neurovascular changes in type 2 diabetes mellitus patients without
clinical diabetic retinopathy Acta diabetol. 55 1075–82
[49] Philip S, Najafi A, Tantraworasin A, Chui T Y P, Rosen R B and Ritch R 2019 Macula
vessel density and foveal avascular zone parameters in exfoliation glaucoma compared to
primary open-angle glaucoma Invest. Ophthalmol. Vis. Sci. 60 1244–53
[50] Alam M, Le D, Lim J I and Yao X Vascular complexity analysis in optical coherence
tomography angiography of diabetic retinopathy Retina 41 538–45
[51] Zahid S et al 2016 Fractal dimensional analysis of optical coherence tomography
angiography in eyes with diabetic retinopathy Invest. Ophthalmol. Vis. Sci. 57 4940–7
[52] Cabral D, Coscas F, Glacet-Bernard A, Pereira T, Geraldes C, Cachado F, Papoila A,
Coscas G and Souied E 2019 Biomarkers of peripheral nonperfusion in retinal venous
occlusions using optical coherence tomography angiography Transl. Vis. Sci. Technol. 8 7
[53] Murray C D 1926 The physiological principle of minimum work: I. The vascular system
and the cost of blood volume Proc. Natl. Acad. Sci. USA 12 207–14
[54] Le D, Alam M, Miao B A, Lim J I and Yao X 2019 Fully automated geometric feature
analysis in optical coherence tomography angiography for objective classification of
diabetic retinopathy Biomed. Opt. Express 10 2493–503
[55] Wang X, Jia Y, Spain R, Potsaid B, Liu J J, Baumann B, Hornegger J, Fujimoto J G, Wu
Q and Huang D 2014 Optical coherence tomography angiography of optic nerve head and
parafovea in multiple sclerosis Br. J. Ophthalmol. 98 1368–73
[56] Zhang Q et al 2018 A novel strategy for quantifying choriocapillaris flow voids using swept-
source OCT angiography quantifying choriocapillaris flow voids using SS-OCTA Invest.
Ophth. Vis. Sci. 59 203–11
[57] Patel R, Wang J, Campbell J P, Kiang L, Lauer A, Flaxel C, Hwang T, Lujan B, Huang D
and Bailey S T 2018 Classification of choroidal neovascularization using projection-
resolved optical coherence tomographic angiography Invest. Ophth. Vis. Sci. 59 4285–91

7-18
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[58] Pechauer A D, Jia Y, Liu L, Gao S S, Jiang C and Huang D 2015 Optical coherence
tomography angiography of peripapillary retinal blood flow response to hyperoxia OCT
angiography of retinal blood flow during hyperoxia Invest. Ophth. Vis. Sci. 56 3287–91
[59] Baumann B, Potsaid B, Kraus M F, Liu J J, Huang D, Hornegger J, Cable A E, Duker J S
and Fujimoto J G 2011 Total retinal blood flow measurement with ultrahigh speed swept
source/Fourier domain OCT Biomed. Opt. Express 2 1539–52
[60] Moult E et al 2014 Ultrahigh-speed swept-source OCT angiography in exudative AMD
Ophthalmic Surg. Lasers Imaging Retina 45 496–505
[61] Ploner S B et al 2016 Toward quantitative optical coherence tomography angiography:
visualizing blood flow speeds in ocular pathology using variable interscan time analysis
Retina 36 S118–26
[62] Uchida A, Hu M, Babiuch A, Srivastava S K, Singh R P, Kaiser P K, Talcott K,
Rachitskaya A and Ehlers J P 2019 Optical coherence tomography angiography character-
istics of choroidal neovascularization requiring varied dosing frequencies in treat-and-
extend management: an analysis of the AVATAR study PLoS One 14 e0218889
[63] Carnevali A, Cicinelli M V, Capuano V, Corvi F, Mazzaferro A, Querques L, Scorcia V,
Souied E H, Bandello F and Querques G 2016 Optical coherence tomography angiography:
a useful tool for diagnosis of treatment-naïve quiescent choroidal neovascularization Am. J.
Ophthalmol. 169 189–98
[64] Alam M, Toslak D, Lim J I and Yao X 2019 OCT feature analysis guided artery-vein
differentiation in OCTA Biomed. Opt. Express 10 2055–66
[65] Son T, Alam M, Kim T-H, Liu C, Toslak D and Yao X 2019 Near infrared oximetry-
guided artery–vein classification in optical coherence tomography angiography Exp. Biol.
Med. 244 813–8
[66] Son T, Alam M, Liu C, Toslak D and Yao X 2019 Optical coherence tomography guided
artery-vein classification in retinal OCT angiography of macular region Ophthalmic
Technologies XXIX (Bellingham, WA: International Society for Optics and Photonics)
108581L
[67] Camino A, Wang Z, Wang J, Pennesi M E, Yang P, Huang D, Li D and Jia Y 2018 Deep
learning for the segmentation of preserved photoreceptors on en face optical coherence
tomography in two inherited retinal diseases Biomed. Opt. Express 9 3092–105
[68] Fang L, Cunefare D, Wang C, Guymer R H, Li S and Farsiu S 2017 Automatic
segmentation of nine retinal layer boundaries in OCT images of non-exudative AMD
patients using deep learning and graph search Biomed. Opt. Express 8 2732–44
[69] Roy A G, Conjeti S, Karri S P K, Sheet D, Katouzian A, Wachinger C and Navab N 2017
ReLayNet: retinal layer and fluid segmentation of macular optical coherence tomography
using fully convolutional networks Biomed. Opt. Express 8 3627–42
[70] Sandhu H S, Eladawi N, Elmogy M, Keynton R, Helmy O, Schaal S and El-Baz A 2018
Automated diabetic retinopathy detection using optical coherence tomography angiogra-
phy: a pilot study Br. J. Ophthalmol. 102 1564–9
[71] Heisler M, Karst S, Lo J, Mammo Z, Yu T, Warner S, Maberley D, Beg M F, Navajas E V
and Sarunic M V 2020 Ensemble deep learning for diabetic retinopathy detection using
optical coherence tomography angiography Transl. Vis. Sci. Technol. 9 20
[72] Le D, Alam M, Yao C K, Lim J I, Hsieh Y-T, Chan R V, Toslak D and Yao X 2020
Transfer learning for automated OCTA detection of diabetic retinopathy Transl. Vis. Sci.
Technol. 9 35

7-19
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[73] Prentašić P, Heisler M, Mammo Z, Lee S, Merkur A, Navajas E, Beg M F, Šarunic M and
Lončarić S 2016 Segmentation of the foveal microvasculature using deep learning networks
J. Biomed. Opt. 21 075008
[74] Guo Y, Camino A, Wang J, Huang D, Hwang T S and Jia Y 2018 MEDnet, a neural
network for automated detection of avascular area in OCT angiography Biomed. Opt.
Express 9 5147–58
[75] Guo Y, Hormel T T, Xiong H, Wang B, Camino A, Wang J, Huang D, Hwang T S and Jia
Y 2019 Development and validation of a deep learning algorithm for distinguishing the
nonperfusion area from signal reduction artifacts on oct angiography Biomed. Opt. Express
10 3257–68
[76] Kasaragod D, Makita S, Hong Y-J and Yasuno Y 2018 Machine-learning based
segmentation of the optic nerve head using multi-contrast Jones matrix optical coherence
tomography with semi-automatic training dataset generation Biomed. Opt. Express 9
3220–43
[77] Alam M, Le D, Lim J I, Chan R V P and Yao X C 2019 Supervised machine learning based
multi-task artificial intelligence classification of retinopathies J. Clin. Med. 8 872
[78] Samara W A, Shahlaee A, Sridhar J, Khan M A, Ho A C and Hsu J 2016 Quantitative
optical coherence tomography angiography features and visual function in eyes with branch
retinal vein occlusion Am. J. Ophthalmol. 166 76–83
[79] Chen J L, Zahid S, Alam M N, Yao X and Lim J I 2018 Assessment of quantitative optical
coherence tomography angiography parameters in branch retinal vein occlusion and
monitoring response to treatment Invest. Ophth. Vis. Sci. 59 5458
[80] Zahid S, Alam M N, Yao X and Lim J I 2018 Quantitative optical coherence tomography
angiography parameters in central retinal vein occlusion Invest. Ophth. Vis. Sci. 59 5427
[81] Ashraf M, Nesper P L, Jampol L M, Yu F and Fawzi A A 2018 Statistical model of optical
coherence tomography angiography parameters that correlate with severity of diabetic
retinopathy Invest. Ophth. Vis. Sci. 59 4292–8
[82] Ouyang Y, Shao Q, Scharf D, Joussen A M and Heussen F M 2014 An easy method to
differentiate retinal arteries from veins by spectral domain optical coherence tomography:
retrospective, observational case series BMC Ophthalmol. 14 66
[83] Joshi V S, Reinhardt J M, Garvin M K and Abramoff M D 2014 Automated method for
identification and artery-venous classification of vessel trees in retinal vessel networks PLoS
One 9 e88061
[84] Zheng F et al 2017 Comparison of neovascular lesion area measurements from different
swept-source OCT angiographic scan patterns in age-related macular degeneration Invest.
Ophthalmol. Vis. Sci. 58 5098–104
[85] de Carlo T E, Romano A, Waheed N K and Duker J S 2015 A review of optical coherence
tomography angiography (OCTA) Int. J. Retin. Vitr. 1 5
[86] Miller A R, Roisman L, Zhang Q, Zheng F, de Oliveira Dias J R, Yehoshua Z, Schaal K B,
Feuer W, Gregori G and Chu Z 2017 Comparison between spectral-domain and swept-
source optical coherence tomography angiographic imaging of choroidal neovasculariza-
tion Invest. Ophth. Vis. Sci. 58 1499–505
[87] Lim H B, Kim Y W, Kim J M, Jo Y J and Kim J Y 2018 The importance of signal strength
in quantitative assessment of retinal vessel density using optical coherence tomography
angiography Sci. Rep. 8 12897

7-20
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[88] Liu L, Gao J, Bao W, Hu C, Xu Y, Zhao B, Zheng J, Fan L and Sun Y 2018 Analysis of
foveal microvascular abnormalities in diabetic retinopathy using optical coherence tomog-
raphy angiography with projection artifact removal J. Ophthalmol. 2018 3926745
[89] Yu S, Frueh B E, Steinmair D, Ebneter A, Wolf S, Zinkernagel M S and Munk M R 2018
Cataract significantly influences quantitative measurements on swept-source optical coher-
ence tomography angiography imaging PLoS One 13 e0204501
[90] Corvi F, Pellegrini M, Erba S, Cozzi M, Staurenghi G and Giani A 2018 Reproducibility of
vessel density, fractal dimension, and foveal avascular zone using 7 different optical
coherence tomography angiography devices Am. J. Ophthalmol. 186 25–31
[91] Tian M, Tappeiner C, Zinkernagel M S, Wolf S and Munk M R 2019 Swept-source optical
coherence tomography angiography reveals vascular changes in intermediate uveitis Acta
Ophthalmol 97 e785–91
[92] Lavia C, Bonnin S, Maule M, Erginay A, Tadayoni R and Gaudric A 2019 Vessel density
of superficial, intermediate, and deep capillary plexuses using optical coherence tomog-
raphy angiography Retina 39 247
[93] Li T, Jia Y, Wang S, Wang A, Gao L, Yang C and Zou H 2019 Retinal microvascular
abnormalities in children with type 1 diabetes mellitus without visual impairment or
diabetic retinopathy Invest. Ophth. Vis. Sci. 60 990–8
[94] Cao D, Yang D, Huang Z, Zeng Y, Wang J, Hu Y and Zhang L 2018 Optical coherence
tomography angiography discerns preclinical diabetic retinopathy in eyes of patients with
type 2 diabetes without clinical diabetic retinopathy Acta Diabetol. 55 469–77
[95] Garrity S T, Iafe N A, Phasukkijwatana N, Chen X and Sarraf D 2017 Quantitative
analysis of three distinct retinal capillary plexuses in healthy eyes using optical coherence
tomography angiography Invest. Ophthalmol. Vis. Sci. 58 5548–55
[96] Fujiwara A, Mengxuan L, Morizane Y, Hosogi M, Kimura S, Hosokawa M, Shiode Y,
Masayuki H, Doi S and Toshima S 2018 Interocular symmetry of foveal avascular zone
area in healthy eyes: an examination using swept-source optical coherence tomography
angiography Invest. Ophth. Vis. Sci. 59 2882
[97] Eastline M, Munk M R, Wolf S, Schaal K B, Ebneter A, Tian M, Giannakaki-
Zimmermann H and Zinkernagel M S 2019 Repeatability of wide-field optical coherence
tomography angiography in normal retina Transl. Vis. Sci. Technol. 8 6
[98] Lutsenko N, Rudycheva O and Isakova O 2019 Assessing OCTA changes in morphology
and structure of retinal microvascular bed in patients with exudative AMD J. Ophthalmol.
(Ukraine) 2 7–13
[99] Agarwal A, Aggarwal K, Akella M, Agrawal R, Khandelwal N, Bansal R, Singh R and
Gupta V 2018 Fractal dimension and optical coherence tomography angiography features of
the central macula after repair of rhegmatogenous retinal detachments Retina 39 2167–77
[100] Richter G M et al 2018 Structural and functional associations of macular microcirculation
in the ganglion cell-inner plexiform layer in glaucoma using optical coherence tomography
angiography J. Glaucoma 27 281–90
[101] Kim S V, Semoun O, Pedinielli A, Jung C, Miere A and Souied E H 2019 Optical coherence
tomography angiography quantitative assessment of exercise-induced variations in retinal
vascular plexa of healthy subjects Invest. Ophth. Vis. Sci. 60 1412–19
[102] Coscas F, Cabral D, Pereira T, Geraldes C, Narotamo H, Miere A, Lupidi M, Sellam A,
Papoila A and Coscas G 2018 Quantitative optical coherence tomography angiography
biomarkers for neovascular age-related macular degeneration in remission PLoS One 13
e0205513

7-21
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[103] Murakawa S, Maruko I, Kawano T, Hasegawa T and Iida T 2019 Choroidal neo-
vascularization imaging using multiple en face optical coherence tomography angiography
image averaging Graefe’s Arch. Clin. Exp. Ophthalmol. 257 1119–25
[104] Dave V P, Pappuru R R, Gindra R, Ananthakrishnan A, Modi S, Trivedi M and
Harikumar P 2018 OCT angiography fractal analysis-based quantification of macular
vascular density in BRVO eyes Can. J. Ophthalmol. 54 297–300
[105] Hirano T, Kitahara J, Toriyama Y, Kasamatsu H, Murata T and Sadda S 2019
Quantifying vascular density and morphology using different swept-source optical coher-
ence tomography angiographic scan patterns in diabetic retinopathy Br. J. Ophthalmol. 103
216–21
[106] Koulisis N, Kim A Y, Chu Z, Shahidzadeh A, Burkemper B, de Koo L C O, Moshfeghi A
A, Ameri H, Puliafito C A and Isozaki V L 2017 Quantitative microvascular analysis of
retinal venous occlusions by spectral domain optical coherence tomography angiography
PLoS One 12 e0176404
[107] Al-Sheikh M, Phasukkijwatana N, Dolz-Marco R, Rahimi M, Iafe N A, Freund K B,
Sadda S R and Sarraf D 2017 Quantitative OCT angiography of the retinal micro-
vasculature and the choriocapillaris in myopic eyes Invest. Ophth. Vis. Sci. 58 2063–9
[108] Kim A Y, Rodger D C, Shahidzadeh A, Chu Z, Koulisis N, Burkemper B, Jiang X, Pepple
K L, Wang R K and Puliafito C A 2016 Quantifying retinal microvascular changes in
uveitis using spectral-domain optical coherence tomography angiography Am. J.
Ophthalmol. 171 101–12
[109] Kim A Y, Chu Z, Shahidzadeh A, Wang R K, Puliafito C A and Kashani A H 2016
Quantifying microvascular density and morphology in diabetic retinopathy using spectral-
domain optical coherence tomography angiography Invest. Ophth. Vis. Sci. 57 OCT362–70
[110] Rosen R B, Krawitz B, Mo S, Geyman L, Phillips E, Carroll J, Weitz R and Chui T Y P
2016 Age-related variations in foveal avascular zone geometry and vessel density-an optical
coherence tomography angiography (OCTA) study x Invest. Ophth. Vis. Sci. 57 5501
[111] Fang D, Tang F Y, Huang H, Cheung C Y and Chen H 2019 Repeatability, interocular
correlation and agreement of quantitative swept-source optical coherence tomography
angiography macular metrics in healthy subjects Br. J. Ophthalmol. 103 415–20
[112] Rabiolo A, Gelormini F, Marchese A, Cicinelli M V, Triolo G, Sacconi R, Querques L,
Bandello F and Querques G 2018 Macular perfusion parameters in different angiocube
sizes: does the size matter in quantitative optical coherence tomography angiography?
Invest. Ophthalmol. Vis. Sci. 59 231–7
[113] Mo S, Krawitz B, Efstathiadis E, Geyman L, Weitz R, Chui T Y, Carroll J, Dubra A and
Rosen R B 2016 Imaging foveal microvasculature: optical coherence tomography angiog-
raphy versus adaptive optics scanning light ophthalmoscope fluorescein angiography Invest.
Ophth. Vis. Sci. 57 OCT130–40
[114] Iafe N A, Phasukkijwatana N, Chen X and Sarraf D 2016 Retinal capillary density and
foveal avascular zone area are age-dependent: quantitative analysis using optical coherence
tomography angiography Invest. Ophth. Vis. Sci. 57 5780–7
[115] Kwon J, Choi J, Shin J W, Lee J and Kook M S 2018 An optical coherence tomography
angiography study of the relationship between foveal avascular zone size and retinal vessel
density Invest. Ophth. Vis. Sci. 59 4143–53
[116] Zhang Y, Do B, Mustafi D, Kogachi K, Chu Z, Wang R K, Rodger D C, Rao N A and
Kashani A H 2018 Assessing response of retinal microvasculature density and morphology

7-22
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

to intra-vitreal dexamethasone implant injections for ocular inflammatory disease with

spectral domain optical coherence tomography angiography (SD-OCTA) Invest. Ophth.
Vis. Sci. 59 2835
[117] Shahidzadeh A, Chen Y C, Chu Z, Wang R K and Kashani A H 2018 Correlation of
macular thickness with microvascular density and morphology using spectral domain
optical coherence tomography angiography (SD-OCTA) among normal subjects Invest.
Ophth. Vis. Sci. 59 2869
[118] Falavarjani K G, Tian J J, Akil H, Garcia G A, Sadda S R and Sadun A A 2016 Swept-
source optical coherence tomography angiography of the optic disk in optic neuropathy
Retina 36 S168–77
[119] Talisa E et al 2015 Detection of microvascular changes in eyes of patients with diabetes but
not clinical diabetic retinopathy using optical coherence tomography angiography Retina
35 2364–70
[120] Lei J, Yi E, Suo Y, Chen C, Xu X, Ding W, Abdelfattah N S, Fan X and Lu H 2018
Distinctive analysis of macular superficial capillaries and large vessels using optical coherence
tomographic angiography in healthy and diabetic eyes Invest. Ophth. Vis. Sci. 59 1937–43
[121] Kaizu Y, Nakao S, Yoshida S, Hayami T, Arima M, Yamaguchi M, Wada I, Hisatomi T,
Ikeda Y and Ishibashi T 2017 Optical coherence tomography angiography reveals spatial
bias of macular capillary dropout in diabetic retinopathy Invest. Ophth. Vis. Sci. 58 4889
[122] Kwon J, Choi J, Shin J W, Lee J and Kook M S 2017 Alterations of the foveal avascular
zone measured by optical coherence tomography angiography in glaucoma patients with
central visual field defects Invest. Ophth. Vis. Sci. 58 1637–45
[123] Kim J-A, Kim T-W, Lee E J, Girard M J and Mari J M 2018 Microvascular changes in
peripapillary and optic nerve head tissues after trabeculectomy in primary open-angle
glaucoma Invest. Ophth. Vis. Sci. 59 4614–21
[124] Muraoka Y, Uji A, Ishikura M, Iida Y, Ooto S and Tsujikawa A 2018 Segmentation of the
four-layered retinal vasculature using high-resolution optical coherence tomography
angiography reveals the microcirculation unit Invest. Ophth. Vis. Sci. 59 5847–53
[125] Anegondi N, Kshirsagar A, Mochi T B and Roy A S 2018 Quantitative comparison of
retinal vascular features in optical coherence tomography angiography images from three
different devices Ophthalmic Sur. Lasers Imaging Retina 49 488–96
[126] Gadde S G, Anegondi N, Bhanushali D, Chidambara L, Yadav N K, Khurana A and Roy
A S 2016 Quantification of vessel density in retinal optical coherence tomography
angiography images using local fractal dimension Invest. Ophth. Vis. Sci. 57 246–52
[127] Wang Q, Chan S, Yang J Y, You B, Wang Y X, Jonas J B and Wei W B 2016 Vascular
density in retina and choriocapillaris as measured by optical coherence tomography
angiography Am. J. Ophthalmol. 168 95–109
[128] Onishi A C, Nesper P L, Roberts P K, Moharram G A, Chai H, Liu L, Jampol L M and
Fawzi A A 2018 Importance of considering the middle capillary plexus on OCT
angiography in diabetic retinopathy Invest. Ophth. Vis. Sci. 59 2167–76
[129] Hwang T S, Gao S S, Liu L, Lauer A K, Bailey S T, Flaxel C J, Wilson D J, Huang D and
Jia Y 2016 Automated quantification of capillary nonperfusion using optical coherence
tomography angiography in diabetic retinopathy JAMA Ophthalmol. 134 367
[130] Jia Y, Bailey S T, Hwang T S, McClintic S M, Gao S S, Pennesi M E, Flaxel C J, Lauer A K,
Wilson D J and Hornegger J 2015 Quantitative optical coherence tomography angiography
of vascular abnormalities in the living human eye Proc. Natl. Acad. Sci. 112 E2395–402

7-23
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[131] Hsieh Y-T, Alam M N, Le D, Hsiao C-C, Yang C-H, Chao D and Yao X 2019 Optical
coherence tomography angiography biomarkers for predicting visual outcomes after
ranibizumab treatment for diabetic macular edema Ophthalmol. Retina 3 826–34
[132] Skalet A et al 2018 Quantitative optical coherence tomography angiography for evaluation
of peripapillary retinal capillary circulation after plaque brachytherapy for uveal melanoma
Ophthalmol. Retina 2 244–50
[133] Chu Z, Lin J, Gao C, Xin C, Zhang Q, Chen C-L, Roisman L, Gregori G, Rosenfeld P J
and Wang R K 2016 Quantitative assessment of the retinal microvasculature using optical
coherence tomography angiography J. Biomed. Opt. 21 066008
[134] Kim T-H, Son T, Lu Y, Alam M and Yao X 2018 Comparative optical coherence
tomography angiography of wild-type and rd10 mouse retinas Transl. Vis. Sci. Technol. 7 42
[135] Kim Y, Hong H K, Park J R, Choi W, Woo S J, Park K H and Oh W-Y 2018 Oxygen-
induced retinopathy and choroidopathy: in vivo longitudinal observation of vascular
changes using OCTA Invest. Ophth. Vis. Sci. 59 3932–42
[136] Inooka D, Ueno S, Kominami T, Sayo A, Okado S, Ito Y and Terasaki H 2018
Quantification of macular microvascular changes in patients with retinitis pigmentosa
using optical coherence tomography angiography Invest. Ophthalmol. Vis. Sci. 59 433–8
[137] Arthur E, Elsner A E, Sapoznik K A, Papay J A, Muller M S and Burns S A 2019 Distances
from capillaries to arterioles or venules measured using OCTA and AOSLO Invest. Ophth.
Vis. Sci. 60 1833–44
[138] Brunner M, Romano V, Steger B, Vinciguerra R, Lawman S, Williams B, Hicks N,
Czanner G, Zheng Y and Willoughby C E 2018 Imaging of corneal neovascularization:
optical coherence tomography angiography and fluorescence angiography Invest. Ophth.
Vis. Sci. 59 1263–9
[139] Hosseinaee Z, Tan B, Martinez A and Bizheva K K 2018 Comparative study of optical
coherence tomography angiography and phase-resolved doppler optical coherence tomog-
raphy for measurement of retinal blood vessels caliber Transl. Vis. Sci. Technol. 7 18
[140] Vujosevic S, Muraca A, Alkabes M, Villani E, Cavarzeran F, Rossetti L and De Cilla S
2019 Early microvascular and neural changes in patients with type 1 and type 2 diabetes
mellitus without clinical signs of diabetic retinopathy Retina 39 435–45
[141] Chu S, Nesper P L, Soetikno B T, Bakri S J and Fawzi A A 2018 Projection-resolved OCT
angiography of microvascular changes in paracentral acute middle maculopathy and acute
macular neuroretinopathy Invest. Ophth. Vis. Sci. 59 2913–22
[142] Dansingani K K and Freund K B 2015 Optical coherence tomography angiography reveals
mature, tangled vascular networks in eyes with neovascular age-related macular degener-
ation showing resistance to geographic atrophy Ophthalmic Sur. Lasers Imaging Retina 46
907–12
[143] Saraf S S, Tyring A J, Chen C-L, Le T P, Kalina R E, Wang R and Chao J R 2019 Familial
retinal arteriolar tortuosity and quantification of vascular tortuosity using swept-source
optical coherence tomography angiography Am. J. Ophthalmol. Case Rep. 14 74–8
[144] Khansari M M, O’Neill W, Lim J and Shahidi M 2017 Method for quantitative assessment
of retinal vessel tortuosity in optical coherence tomography angiography applied to sickle
cell retinopathy Biomed. Opt. Express 8 3796–806
[145] Bulut M, Kurtuluş F, Gözkaya O, Erol M K, Cengiz A, Akıdan M and Yaman A 2018
Evaluation of optical coherence tomography angiographic findings in Alzheimer’s type
dementia Br. J. Ophthalmol. 102 233–37

7-24
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[146] Chang M Y, Phasukkijwatana N, Garrity S, Pineles S L, Rahimi M, Sarraf D, Johnston M,

Charles A and Arnold A C 2017 Foveal and peripapillary vascular decrement in migraine
with aura demonstrated by optical coherence tomography angiography Invest. Ophth. Vis.
Sci. 58 5477–84
[147] Conti F F, Qin V L, Rodrigues E B, Sharma S, Rachitskaya A V, Ehlers J P and Singh R P
2019 Choriocapillaris and retinal vascular plexus density of diabetic eyes using split-
spectrum amplitude decorrelation spectral-domain optical coherence tomography angiog-
raphy Br. J. Ophthalmol. 103 452–6
[148] Dimitrova G, Chihara E, Takahashi H, Amano H and Okazaki K 2017 Quantitative retinal
optical coherence tomography angiography in patients with diabetes without diabetic
retinopathy Invest. Ophth. Vis. Sci. 58 190–6
[149] Garrity S T, Iafe N A, Phasukkijwatana N, Chen X and Sarraf D 2017 Quantitative
analysis of retinal capillary density and foveal avascular zone area using optical coherence
tomography angiography of normal eyes Invest. Ophth. Vis. Sci. 58 3646
[150] Goudot M M, Sikorav A, Semoun O, Miere A, Jung C, Courbebaisse B, Srour M, Freiha J
G and Souied E H 2017 Parafoveal OCT angiography features in diabetic patients without
clinical diabetic retinopathy: a qualitative and quantitative analysis J. Ophthalmol. 2017
8676091
[151] Kim K, Kim E S and Yu S-Y 2018 Optical coherence tomography angiography analysis of
foveal microvascular changes and inner retinal layer thinning in patients with diabetes Br. J.
Ophthalmol. 102 1226–31
[152] Linderman R, Salmon A E, Strampe M, Russillo M, Khan J and Carroll J 2017 Assessing
the accuracy of foveal avascular zone measurements using optical coherence tomography
angiography: segmentation and scaling Transl. Vis. Sci. Technol. 6 16
[153] Samara W A, Shahlaee A, Adam M K, Khan M A, Chiang A, Maguire J I, Hsu J and Ho
A C 2017 Quantification of diabetic macular ischemia using optical coherence tomography
angiography and its relationship with visual acuity Ophthalmology 124 235–44
[154] Scarinci F, Picconi F, Giorno P, Boccassini B, De Geronimo D, Varano M, Frontoni S and
Parravano M 2018 Deep capillary plexus impairment in patients with type 1 diabetes
mellitus with no signs of diabetic retinopathy revealed using optical coherence tomography
angiography Acta Ophthalmol. 96 e264–5
[155] Simonett J M, Scarinci F, Picconi F, Giorno P, De Geronimo D, Di Renzo A, Varano M,
Frontoni S and Parravano M 2017 Early microvascular retinal changes in optical coherence
tomography angiography in patients with type 1 diabetes mellitus Acta Ophthalmol. 95 e751–5
[156] Yeung L, Wu I W, Sun C C, Liu C F, Chen S Y, Tseng C H, Lee H C and Lee C C 2019
Early retinal microvascular abnormalities in patients with chronic kidney disease
Microcirculation e12555
[157] Yoon Y S, Woo J M, Woo J E and Min J K 2018 Superficial foveal avascular zone
area changes before and after idiopathic epiretinal membrane surgery Int. J. Ophthalmol.
11 1711
[158] Alibhai A Y, Moult E M, Shahzad R, Rebhun C B, Moreira-Neto C, McGowan M, Lee D,
Lee B, Baumal C R and Witkin A J 2018 Quantifying microvascular changes using OCT
angiography in diabetic eyes without clinical evidence of retinopathy Ophthalmol. Retina 2
418–27
[159] Andrade Romo J S, Linderman R E, Pinhas A, Carroll J, Rosen R B and Chui T Y P 2019
Novel development of parafoveal capillary density deviation mapping using an age-group and
eccentricity matched normative OCT angiography database Transl. Vis. Sci. Technol. 8 1

7-25
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[160] Hwang T S, Jia Y, Gao S S, Bailey S T, Lauer A K, Flaxel C J, Wilson D J and Huang D
2015 Optical coherence tomography angiography features of diabetic retinopathy Retina
35 2371
[161] Kaizu Y, Nakao S, Wada I, Yamaguchi M, Fujiwara K, Yoshida S, Hisatomi T, Ikeda Y,
Hayami T and Ishibashi T 2017 Imaging of retinal vascular layers: adaptive optics scanning
laser ophthalmoscopy versus optical coherence tomography angiography Transl. Vis. Sci.
Technol. 6 2
[162] Krawitz B D, Phillips E, Bavier R D, Mo S, Carroll J, Rosen R B and Chui T Y 2018
Parafoveal nonperfusion analysis in diabetic retinopathy using optical coherence tomog-
raphy angiography Transl. Vis. Sci. Technol. 7 4
[163] Lu Y, Simonett J M, Wang J, Zhang M, Hwang T, Hagag A M, Huang D, Li D and Jia Y
2018 Evaluation of automatically quantified foveal avascular zone metrics for diagnosis of
diabetic retinopathy using optical coherence tomography angiography Invest. Ophth. Vis.
Sci. 59 2212–21
[164] Philip S, Najafi A, Tantraworasin A, Chui T Y, Rosen R B and Ritch R 2019 Macula vessel
density and foveal avascular zone parameters in exfoliation glaucoma compared to primary
open-angle glaucoma Invest. Ophth. Vis. Sci. 60 1244–53
[165] Rosen R B, Andrade Romo J S, Krawitz B D, Mo S, Fawzi A A, Linderman R E, Carroll J,
Pinhas A and Chui T Y P 2019 Earliest evidence of preclinical diabetic retinopathy
revealed using optical coherence tomography angiography perfused capillary density
Am. J. Ophthalmol. 203 P103–15
[166] Samara W A, Say E A, Khoo C T, Higgins T P, Magrath G, Ferenczy S and Shields C L
2015 Correlation of foveal avascular zone size with foveal morphology in normal eyes using
optical coherence tomography angiography Retina 35 2188–95
[167] Schmidt T G, Linderman R E, Strampe M R, Chui T Y P, Rosen R B and Carroll J 2019
The utility of frame averaging for automated algorithms in analyzing retinal vascular
biomarkers in angioVue OCTA Transl. Vis. Sci. Technol. 8 10
[168] Takase N, Nozaki M, Kato A, Ozeki H, Yoshida M and Ogura Y 2015 Enlargement of
foveal avascular zone in diabetic eyes evaluated by en face optical coherence tomography
angiography Retina 35 2377–83
[169] Tan C S, Lim L W, Chow V S, Chay I W, Tan S, Cheong K X, Tan G T and Sadda S R
2016 Optical coherence tomography angiography evaluation of the parafoveal vasculature
and its relationship with ocular factors Invest. Ophthalmol. Vis. Sci. 57 OCT224–234
[170] Anegondi N, Chidambara L, Bhanushali D, Gadde S G, Yadav N K and Sinha Roy A
2018 An automated framework to quantify areas of regional ischemia in retinal vascular
diseases with OCT angiography J. Biophotonics 11 e201600312
[171] Sioufi K, Shahlaee A, Say E A T, Ferenczy S, Hsu J and Shields C L 2017 Fractal
dimension analysis of parafoveal microvascular anatomy using optical coherence tomog-
raphy angiography in two machines Invest. Ophth. Vis. Sci. 58 1674
[172] Faatz H, Farecki M-L, Rothaus K, Gunnemann F, Gutfleisch M, Lommatzsch A and
Pauleikhoff D 2019 Optical coherence tomography angiography of types 1 and 2 choroidal
neovascularization in age-related macular degeneration during anti-VEGF therapy: eval-
uation of a new quantitative method Eye 33 1466–71
[173] Ting D S W, Tan G S W, Agrawal R, Yanagi Y, Sie N M, Wong C W, San Yeo I Y, Lee S Y,
Cheung C M G and Wong T Y 2017 Optical coherence tomographic angiography in type 2
diabetes and diabetic retinopathy JAMA Ophthalmol. 135 306–12

7-26
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[174] Chen Q, Ma Q, Wu C, Tan F, Chen F, Wu Q, Zhou R, Zhuang X, Lu F and Qu J 2017

Macular vascular fractal dimension in the deep capillary layer as an early indicator of
microvascular loss for retinopathy in type 2 diabetic patients Invest. Ophth. Vis. Sci. 58
3785–94
[175] Bhardwaj S, Tsui E, Zahid S, Young E, Mehta N, Agemy S, Garcia P, Rosen R B and
Young J A 2018 Value of fractal analysis of optical coherence tomography angiography in
various stages of diabetic retinopathy Retina 38 1816–23
[176] Spain R I, Liu L, Zhang X, Jia Y, Tan O, Bourdette D and Huang D 2018 Optical
coherence tomography angiography enhances the detection of optic nerve damage in
multiple sclerosis Br. J. Ophthalmol. 102 520–4
[177] Ishibazawa A, Nagaoka T, Takahashi A, Omae T, Tani T, Sogawa K, Yokota H and
Yoshida A 2015 Optical coherence tomography angiography in diabetic retinopathy: a
prospective pilot study Am. J. Ophthalmol. 160 35–44 e31
[178] Rebhun C B et al 2017 Polypoidal choroidal vasculopathy on swept-source optical
coherence tomography angiography with variable interscan time analysis Transl. Vis. Sci.
Technol. 6 4
[179] Talisa E, Bonini Filho M A, Chin A T, Adhi M, Ferrara D, Baumal C R, Witkin A J,
Reichel E, Duker J S and Waheed N K 2015 Spectral-domain optical coherence
tomography angiography of choroidal neovascularization Ophthalmology 122 1228–38
[180] Kermany D S, Goldbaum M, Cai W, Valentim C C, Liang H, Baxter S L, McKeown A,
Yang G, Wu X and Yan F 2018 Identifying medical diagnoses and treatable diseases by
image-based deep learning Cell 172 1122–31 e1129

7-27
 IOP Publishing

Photo Acoustic and Optical Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
Ayman El-Baz and Jasjit S Suri

Chapter 8
Clinical utility of OCT-angio in age-related
macular degeneration
Hashim Ali Khan, Smaha Jahangir and Julie Friedman Rodman

Age-related macular denegation (AMD) is a leading cause of blindness among the

elderly in developed countries. Beginning as an outer retinal and choriocapillaris (CC)
metabolical imbalance, it progresses into one of the two advanced forms, i.e.,
geographic atrophy or choroidal neovascularization. Drusen are an early sign of
AMD. Optical coherence tomography (OCT) has been used for evaluation and
documentation of changes associated with drusen, pigment epithelial detachments,
choroidal neovascularization as well as being helpful in decision making. A significant
number of eyes develop choroidal neovascularization and require frequent intra-
ocular injections of vascular endothelial growth factor inhibitors. Label based
angiographic techniques are typically used in diagnosing and management of such
cases while most practices supplement the angiographic studies with OCT.
OCT-angio (OCTA) is evolving as an indispensable tool in providing high resolution,
three-dimensional vascular flow maps as well as structural information of choroidal
retinal tissue; thus providing important biomarkers in diagnosing and guiding the
treatment decision. The diagnostic accuracy of OCTA is comparable to or better (in
some instances) than the current, invasive vascular imaging techniques.
Detailed visualization of the CC was not achievable with other vascular imaging
modalities. OCTA, offers the details of the CC in high resolution. Besides its role in
clinical management of AMD, it has expanded our cognizance about the role of the
CC in pathogenesis and progression of the disease.
In this chapter, we describe clinical as well as OCTA features of different lesions
associated with AMD. In addition, recent classification and lexicons for describing
these entities are discussed.
AMD is the leading cause of irreversible visual loss in the elderly and its
prevalence is estimated to double by 2050. It is derived from several genetic and

doi:10.1088/978-0-7503-2060-3ch8 8-1 ª IOP Publishing Ltd 2021

 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

modifiable environmental factors. AMD can be categorized into early and late
stages. Early and intermediate AMD is characterized by good visual functions, focal
sub-retinal pigment epithelium (RPE) deposits and a variable degree of pigmentary
changes. Late-stage AMD assumes one of the two prevalent forms as choroidal
neovascularization (CNV) and geographic atrophy (GA).
Photoreceptors, RPE, Bruch’s membrane (BM) and the CC are involved
subclinically earlier in the disease course. While one of the entities associated with
AMD (end-stage disease) is geographic atrophy, most eyes with CNV, if untreated,
rapidly progress to disciform scarring of the macula. Early diagnosis and treatment
of the disease may help avoid many complications and slow down the progression
significantly.
OCT is the standard imaging modality for investigating the structural changes of
retinal tissue while fundus fluorescein angiography (FFA) and indocyanine green
angiography (ICGA) are used for evaluation of retinal and choroidal vasculature.
Often, both OCT and FFA are needed to diagnose AMD and differentiate between
the types. However, ICGA is frequently needed because the details of chroidal
vascular changes are not well exhibited by FFA, specifically in occult membranes
that reside below the RPE; because fluorescein dye does not penetrate past the blood
retinal barrier into the choroid.
OCTA, as mentioned elsewhere in this book, can render depth-resolved, high
resolution retinal and choroidal vascular flow maps along with structural informa-
tion. Side-by-side study of retinal morphology and flow characteristics increases our
diagnostic capabilities and accuracy several fold and is very helpful in the early
diagnosis and differentiation from other masquerading clinical entities.

8.1 Clinical features and OCTA correlates of early AMD

Drusen are the hallmark clinicopathological feature of dry AMD. The term dry
emphasizes the absence of fluid around the neurosensory retina that is typical in the
wet form of the disease.
Multiple variants of drusen are described in dry AMD. Nonetheless, drusen-like
lesions are seen in other retinal conditions also. Clinicians must be able to differ-
entiate between the types of drusen and comparable retinal lesions for appropriate
management of these cases. Structural OCT is used for depth localization and
measurement, classification and to evaluate the associated changes in morphology
within the photorecptors and adjacent layers.
Hard drusen, also known as drupelets, are well defined, small, yellow-white
deposits between the RPE and BM. Their size is less than 63 μm. Hard drusen, on
OCT, are seen as variably hyperreflective, small sub-RPE deposits while soft drusen
assume a mound-shaped, hyperreflective elevation of RPE of variable size.
Soft drusen are larger in size (> 63 μm), yellowish-white deposits with indistinct
margins. Often the adjacent soft drusen coalesce to form confluent soft drusen also
known as drusenoid pigment epithelial detachment (PEDs) (figure 8.1). Drusenoid
PEDs are difficult to distinguish from large soft drusens clinically as well as on OCT.

8-2
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 8.1. Shows OCT and OCTA of confluent soft drusen. Normal vascular density in superficial and deep
slabs and no abnormal flow signal in the outer retina. Multiple hyperreflective RPE deposits on OCT cross-
sections (bottom panels) and corresponding dark spots (false low flow versus low flow) on CC flow map
(top right).

Figure 8.2. Multiple drusenoid PEDs with moderate internal reflectivity on cross sections (bottom panels) with
normal retinal flow maps and a variable degree of underlying CC shadowing or flow deficits (top right panel).

They are associated with an increased risk of CNV. There may be an associated
shallow serous retinal detachment in between the large drusens (figure 8.2).
Cuticular or basal laminar drusen are seen as multiple, uniform, yellowish-white,
round punctate densely packed sub RPE deposits representing the nodular excres-
cences of BM. They are better characterized by fluorescein angiography (FA) as
numerous hyperfluorescent foci taking on the appearance of the starry sky.
Cuticular drusens on OCT are prolate shaped, moderately reflective deposits that
usually erode the RPE monolayer assuming a sawtooth appearance. These are
associated with an increased risk of CNV and pseudo-vitelliform macular detach-
ment. The regression of drusen leads to reversal of RPE changes, however, it is an
indicator of progression to advanced disease (i.e. geographic atrophy).

8-3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Reticular pseudodrusen (RPD) are characterized by an interlacing pattern of

subretinal yellowish deposits. Unlike other types of drusen, they are not hyper-
fluorescent on FA. Suzuki et al [1] described three distinct morphological types of
RPD on OCT. Dot pseudodrusen are distributed in the perifovea as discrete whitish
accumulations. The second type termed as ribbon pseudodrusen are characterized
by interconnected bands of yellowish-white material in a reticular pattern.
The third uncommon type, termed as peripheral pseudodrusen, is defined by the
presence of yellow-white globules located outside the major vascular arcades [2–4]
Querques et al [5] described OCT-based staging for the RPD. Granular hyper-
reflective material between the RPE and the ellipsoid zone (EZ) is defined as stage I.
Stage II RPD are mounds of hyperreflective material with alterations in EZ band
contour; that progresses to a thicker hyperreflective material with a conical
morphology eroding through EZ in stage III. The material fades and resorbs/
migrates into the inner retina in stage IV. Light stripes in choroid are observed under
a subset of RPD corresponding to areas of structural damage in the deeper choroid
in these eyes. However, this phenomenon is independent of the size of these deposits
(figure 8.3).

8.1.1 Role of chorioretinal flow characteristic in pathogenesis of drusen

With the advent of OCTA, much attention has been paid to the role of the CC in the
pathogenesis of AMD. While OCTA is becoming an indespensible tool for diagnosis
and management of choroidal neovascularization; the relationship between drusen
and chorioretinal flow characteristics on OCTA has been studied by numerous
authors. CC flow deficit underneath drusen is universal. However, it is debated if
drusen are associated with true hypoflow or false hypoflow signals induced by
shadowing or signal attenuation caused by drusen [6]. Vascular flow information
acquired with spectral-domain (SD) OCTA devices is more susceptible to shadowing
and signal attenuation, giving rise to false hypoflow signals. Swept-source (SS)
OCTA has enhanced tissue penetration and is least prone to shadowing and
masking, yet a possible pseudo flow impairment can not be ruled out.
Nonetheless, the true flow impairment was demonstrated in the CC under drusen
using variable interscan time analysis (VISTA) on SS-OCTA devices [7–9]. CC flow
impairment may be a result of reduced flow speeds (slower than the decorrelation
speed), capillary dropout, or reduced vascular caliber. All these findings are
consistent with past histological evidence about CC hypoxia being a major driver

Figure 8.3. Structural OCT of soft and reticular pseudo-drusen.

8-4
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

behind the metabolic compromise of RPE, BM and outer retina with consequent
drusen formation.
Clear evidence exists supporting the CC flow characteristic as surrogate to RPE
and outer retinal metabolism. Several studies have confirmed drusen formation at
the site of previous CC compromise and there exists a close relationship between CC
insufficiency related to GA development and progression of GA.
Besides drusenoid PEDs, RPD and cuticular drusen are associated with increased
risk of CNV and GA. OCTA may help in diagnosing CNV earlier in the course of
the disease and result in better visual outcomes.

8.2 Geographic atrophy

Geographic atrophy is one form of advanced dry AMD characterized by a well-
demarcated, round or oval area of outer retinal and RPE atrophy seen as a
hypopigmented/depigmented lesion with enhanced visibility of the underlying
choroidal vasculature, that gradually progresses. The fovea may be spared initially
and involved later.
A number of imaging strategies including color fundus images, fundus auto-
fluorescence (AF), multicolor confocal scanning laser ophthalmoscopy (CSLO)
and structural OCT are used to classify the geographic atrophy and document
progression.
Two prevalent forms of GA are drusen associated geographic atrophy (DAGA)
and nascent geographic atrophy (nGA). DAGA is characterized by overt atrophy of
RPE and outer retina while nGA is subsidence of retinal tissue from outer plexiform
layer (OPL) to inner nuclear layer (INL) on OCT. However, the classification of
atrophy meeting (CAM) group has devised a consensus statement on OCT-based
definition and grading of AMD-associated macular atrophy. The group proposed
four categories as complete RPE and outer retinal atrophy (cRORA), incomplete
RPE and outer retinal atrophy (iRORA), complete outer retinal atrophy (cORA),
and incomplete outer retinal atrophy (iORA) [10] (table 8.1).
OCTA has furthered our insights into the pathogenesis of dry AMD. Many
studies have verified the ability of OCTA to detect CC flow impairment under and
beyond the GA margins and have postulated CC flow compromise as a key factor in
the progression and development of GA. GA develops in the outer retina
corresponding to the area of previous CC flow impairment. Considering the CC
flow impairment beyond GA margins, investigators now advocate the use of OCTA
for delineating the extent of atrophy and to monitor progression [11].
Nonetheless, vascular rarefactions reportedly appear initially in the deep capillary
plexus (DCP) and then in the superficial vascular plexus (SVP) even at intermediate
stages of AMD.
It is worth mentioning that OCTA images in drusen, as well as GA, are prone to
artifacts inherent to OCT as well as OCTA. The details of these artifacts have been
mentioned elsewhere in this book. The clinician should take these artifacts into
account while interpreting OCTA images.

8-5
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 8.1. Description and schematic of normal and all categories of macular atrophy on OCT.
Normal macular Normal
morphology. All retinal
layers are well defined
and visible.

Complete homogenous cho- cRORA

roidal hyper-transmission.
Complete loss of RPE,
photoreceptors, EZ band,
ELM and absence of ONL.

Discontinuous choroidal iRORA

hyper-transmission.
Discontinuous RPE band.
Interrupted photoreceptor
layer, EZ band and ELM.
Subsidence of INL & ONL.

None to intermittent hyper- cORA

transmission.
Intact RPE band, photore-
ceptors, EZ band and ELM
are absent.
Sever thinning and inter-
ruption of ONL.

No choroidal hyper- iORA

transmission.
Intact RPE and ELM
bands.
EZ may be disrupted in
presence of regressing SDD.
Detectable thinning of
ONL.
cRORA = complete RPE and outer retinal atrophy, iRORA = incomplete RPE and outer retinal atrophy,
cORA = complete outer retinal atrophy, iORA = incomplete outer retinal atrophy, RPE = retinal pigment
epithelium, ELM = external limiting membrane, EZ = ellipsoid zone, INL = inner nuclear layer, ONL = outer
nuclear layer. RNFL: retinal nerve fiber layer, GCL: ganglion cell layer, OPL: outer plexiform layer, IPL: inner
plexiform layer, BM: Bruch’s membrane, CH: choroidal hyper-transmission, SDD: subretinal drusenoid deposits..

8.3 Neovascular AMD

Choroidal neovascularization (CNV), now called macular neovascularization
(MNV), is characterized by the formation of aberrant vascular proliferation under
the RPE or sometimes above the RPE, which is the other end of AMD. Although
most of the neovascularization in neovascular AMD (nvAMD) originates in the

8-6
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

choroid, a significant number of neovascular membranes originate in intraretinal

tissue. Consensus from the neovascular age-related macular degeneration nomen-
clature (CONAN) group has recently developed a lexicon for standard terminology
in nvAMD. We are using the standard terms described by the group and describe the
older titles for these entities.

8.3.1 Type 1 MNV

Previously known as occult CNV or type 1 CNV, comprises all kinds of neo-
vascularization that originate in the CC and remain confined to the sub-RPE space
and associated with varying degrees of PED. Lesion growth occurs concurrently with
remodeling and broadening of feeder vessels both within the lesion and choroid [12].
Based on ill-defined areas of multipunctate hyperfluorescence/leakage on FA, the
term occult CNV was coined to describe this subtype. ICGA offers relatively better
visualization of these lesions as early hyperfluorescent spots, plaque-like hyper-
fluorescence, and later hyperfluorescence with indistinct margins. Yet the fine details
of neovascular complex cannot be visualized using ICGA.
On structural OCT, this MNV subtype is associated with a PED in 98% of cases
either as incomplete RPE elevation (in up to 35% cases) or complete RPE detach-
ment (in 63%) with accompanying subretinal fluid and intraretinal fluid, in retinal
cystic changes. A double-layer sign, i.e., two hyperreflective bands indicating an
inner RPE and outer BM band separated by a hyporeflective space, is often seen.
The double-layer sign denotes a shallow sub-RPE lesion associated with non-
exudative neovascular membranes [12, 13].
Unlike ICGA and FA, OCTA clearly displays a distinct pattern of a well-
organized vascular network in CC slabs, with evident feeder trunks in most cases.
Type 1 MNV takes one of the three appearances on CC segmentation:
• Medusa pattern: a central trunk vessel with thin branches radiating
centrifugally.
• Sea-fan pattern: an eccentric trunk vessel with most of the neovessels
radiating to one side
• Indistinct pattern: there is no distinguishable pattern and a feeder vessel is not
visible.

Most type 1 MNVs assume a medusa pattern of clear branching vascular network
or larger caliber, more mature vessels confined to sub-RPE space [6, 14–16]
(figure 8.4).
In up to 70% of cases, type 1 neovascular complexes grow significantly under
RPE without causing visual symptoms and anatomical alterations. These neo-
vascular membranes were termed as quiescent CNV. However, the nomenclature
group has suggested calling it type 1 MNV omitting the term quiescent [6, 14–16].
Polypoidal choroidal vasculopathy (PCV) is another variant of wet AMD,
accounting for half of the macular neovascularization in the Asian population. It
is characterized by subretinal neovascularization seen as branching vascular network
(BVN) with nodular dilatations of vessels called polyps. It is best depicted on ICGA

8-7
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 8.4. Showing hyperreflective lesion under RPE on OCT (bottom panel) and corresponding network of
wide caliber vessels in CC slabs (top right panel) and normal outer retinal flow map (top left) on OCTA.
Double-layer sign is evident on OCT.

as initial filling of BVN followed by staining of polyps. The neovascular complex

gradually progresses underneath the Sub-RPE space to a significant size before
causing visual symptoms. Polyps are predisposed to bleed and frequently cause
hemorrhagic (serosanguinous PEDs) PED in these eyes. Although BVN can be
clearly visualized on CC slabs on OCTA, the slow flow in polyps cannot produce a
decorrelation of sufficient strength and are not detected. However, most studies
evaluating polypoidal lesions studied the CC slabs while polyps are usually located
at the top of PED and not segmented on default CC slabs but are better visualized
on avascular outer retinal (AOR) segmentation [17, 18]. Shortly after the initial
description of PCV as polyps, histological studies evidenced their origin from
outpouching/ outgrowth of cavernous thin-walled vasculature underneath RPE.
Considering the histological data, some authors called it aneurysmal type 1
neovascularization. But, the CONAN group classifies polypoidal choroidal neo-
vasculopathy also as type 1 MNV because the neovascular complex does not break
RPE and has its origin from choroidal vasculature (figure 8.5).

8-8
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 8.5. Another example of type 1 MNV.

8.3.2 Type 2 MNV

The type 2 neovascularization originates in the CC and traverses the RPE/BM
complex to grow into the subretinal space. Despite the neovascular process
beginning in CC, the pre-epithelial component leads to exudation and hemorrhaging
in the outer retina with resultant visual symptoms and anatomical changes in retinal
tissue. A well-defined, hyperfluorescent vascular lesion in the early phase of FFA
with subsequent progressive leakage is typical of type 2 MNV [12].
Structural OCT has long been used in diagnosing the disease and guiding
treatment decisions in clinical as well as research settings. On OCT, type 2
neovascular membranes are ascertained as the hyperreflective area between the
RPE/BM and outer retina with varying degrees of associated posterior shadowing.
Disease activity is described by the presence of subretinal and intraretinal fluid, with
cystic changes leading to alterations in macular morphology.
OCTA exhibits one of the two morphological patterns (i.e. medusa or glomerular)
on AOR slabs and the CC as a well-organized network of vessels giving rise to
hyperflow signals. Often one or more, large caliber feeder trunks are seen
anastomosing the MNV to deeper choroidal vasculature. Compared to other vessels,

8-9
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 8.6. OCT of type 2 MNV (E) showing a hyperreflective mass with overlying cystic changes. Hyperflow
signal corresponding to neovascular complex can be seen in the outer retina (C) and CC (D).

trunk vessels are larger in diameter. They sprout and give rise to smaller vessels of
neovascular complex but have minimal branching and choroidal prominent anas-
tomosis with larger vessels [15]. A perilesional dark halo is often seen which
corresponds to CC flow deficit around the lesion (figure 8.6).
The hypoxic environment resulting from CC compromise triggers the release of
vascular endothelial growth factor (VEGF) and other angiogenic growth factors from
overlying RPE cells giving rise to MNV [6, 9,14–16,19–21]. It can be assumed that MNV
develops at the site of previous/longstanding flow deficit and grows in areas correspond-
ing to CC flow loss. Longitudinal studies may be needed to study this assumption.

8.3.3 Type 3 MNV

Type 3 neovascularization, previously known as retinal angiomatous proliferation,
occult chorioretinal anastomosis, retinal vascular anomalous complex and retinal
choroidal anastomosis, originates in the retinal DCP and extends toward the
subretinal space forming retinal-retinal anastomosis and subsequently giving rise
to a connected subretinal neovascular complex followed by proliferation past RPE.
The neovascular complex is prone to leak and bleed causing intraretinal hemor-
rhages, edema and cystic changes. Over time, retinal vasculature supplying the
neovascular component undergoes telangiectatic remodeling to compensate flow
requirement. Although there may be a stimulating initial or concurrent choroidal
component, the main feature is proliferation within the retina [22].
During the intraretinal stage, focal area of intraretinal staining with indistinct
margin is seen on FFA corresponding to the focal intense hyperfluorescent hot spot
on ICGA. A vascularized PED is often seen. Due to progressive leakage and retinal
edema, visualization of the anastomosis is obscured on FA and ICGA [12].

8-10
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

OCT is used to study the morphological changes associated with MNV. A

hyperreflective intraretinal mass is the OCT correlate of the intraretinal component.
It is also helpful in documenting the monitoring of retinal edema and differentiating
vascularized PEDs from serous [22].
Experience with OCTA in type 3 MNV is limited. It takes a distinctive
appearance on OCTA as a small tuft of high flow small-caliber vessels in AOR
segmentation originating from DCP. The retinal-retinal anastomosis can be
identified on OCTA. Abnormal hyperflow signals can also be appreciated in eyes
with PED and sub-RPE proliferation [6, 16, 23, 24]. Tan et al [24] using OCTA and
co-registered OCT Bscans, described two distinct flow patterns in the eye with type 3
MNV. Abnormal hyperflow signal was limited to the neurosensory retina in up to
11% of eyes while in the rest the flow signal extended through the RPE. In a study by
Miere et al [25], OCTA was able to demonstrate type 3 MNV in all 18 eyes as a
distinct tuft of hyperflow-vascular network in AOR abutting posteriorly in sub-RPE
space and anteriorly anastomosing to DCP. Also, a small clew-like hyperflow lesion
was seen in 15 eyes anastomosing to choroid in two eyes [25] (figure 8.7).
OCTA has furthered our understanding of the pathogenic mechanisms associated with
different types of MNVs. As discussed above, both type 1 and type 2 MNVs originate in
the CC and later proliferate past the RPE into the subretinal space; both demonstrate
different features on CC segmentations. Contrary to type 2 lesions, the demarcation
between type 1 neovascularization and adjacent CC is not well-defined. Compared to
type 2, Type 1 MNVs, as evidenced by OCTA, have greater linear dimensions and wider
vascular caliber. The trunk vessels supplying type 1 MNVs have larger caliber than those
of type 2 MNVs. These data establish that type 1 MNVs bear more mature vessels which
are resistant to treatment with antiVEGF drugs. Type 1 MNVs grow up significantly
under RPE before leading to symptoms and remain undetected for long. Formation of
these membranes is in in-part driven by VEGF during angiogenesis and later by other
growth factors through arteriogenesis and vascular normalization [15].

8.4 Diagnostic accuracy of OCTA

Multiple studies have compared OCTA with FA and ICGA for detection of MNV
demonstrating diagnostic accuracy comparable to FA or ICGA. However, being
non-invasive, quick and supplemented by structural information of chorioretinal
tissue, OCTA is more advantageous when compared to FA or ICGA alone [6, 26].
Wang et al [27] performed a metanalysis of 16 studies reporting the diagnostic
accuracy of OCTA compared to either FA or ICGA. They reported a pooled
sensitivity and specificity as 0.87 and 0.96. Others have reported sensitivities as low
as 50% and as high as 90% [14, 28–30]. Nevertheless, en face OCTA flow maps have
diagnostic capabilities comparable to either FFA or ICGA but when combined with
structural OCT and cross-sectional flow maps OCTA had greater diagnostic
accuracy compared to either FFA or ICGA alone.
Most of the MNVs are associated with macular edema; and leakage of dye makes it
difficult to ascertain vascular features of neovascular complexes using dye-based
vascular imaging modalities. However on OCTA, vascular flowmap extraction is not

8-11
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 8.7. Type 3 MNV on color fundus images (a and b). Abnormal hyperflow signal neovascular network
in outer retina (e) and CC (f). Arrow points to the anastomosing trunk from deep capillary plexus (d) with
outer retina.

affected significantly and detailed visualization of neovascular complexes and associ-

ated vascular changes is more readily visible. Furthermore, early MNVs associated with
drusenoid or other types of small PEDs can be missed on FA but a careful evaluation of
cross-sectional and enface flowmaps can be very helpful in detecting them early.

8-12
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

While interpreting OCTA images from different platforms or comparing them to

dye based imaging methods, it must be taken into account that the MNV sizes
measured using SD-OCTA are significantly smaller than those acquired using SS-OCT
and dye-based methods [14].
Many factors influence the diagnostic accuracy of OCTA in detecting MNV;
signal strength, retinal architecture changes, pigmentary abnormalities, and artifacts
are most important. It is important to look and correct for segmentation artifacts
while grading the OCTA image. Structural information from OCT B scans in addition
to en face OCTA flow maps improves the detection of MNVs significantly [31]. (see the
detailed discussion elsewhere in this book).

8.5 OCTA for guiding treatment of AMD

FA and ICGA are gold-standard imaging modalities for the diagnosis of macular
neovascularization. Nonetheless, with the advent of anti-VEGF injections for
neovascular AMD, structural OCT has become a standard for monitoring the
anatomical changes as a response to treatment. The decision to treat and monitor is
partially guided by retinal thickness measured by OCT. As frequent injections are
needed to eradicate neovascular activity, clinicians may need to acquire FA or
ICGA repeatedly for studying the impact of treatment on the neovascular mem-
branes. Owing to the invasive nature of both of these tests, there is a risk of
associated adverse events which increases with every repetition of the tests.
It is known that OCTA cannot detect the neovascular complexes and membranes
with low flow or no flow, and FFA or ICGA demonstrate better sensitivity for such
neovascular membranes. However, we believe that MNVs with no flow or slow flow
are fibrosed complexes which do not respond better to treatment and are usually
monitored [6, 32]. OCTA can be used for monitoring these membranes for both
structural changes (exudation, retinal thickness changes, morphological variations)
and for changes in flow characteristics that also aids in making treatment decisions.
OCTA can provide flow information both as cross-sectional and depth-resolved
en face images at any desired depth in chorioretinal tissue. Multiple objective metrics
are available on commercial devices which can be used to gauge the treatment effects
on the neovascular complex as well as retinal morphology. MNV flow area, vascular
density and retinal thickness are automatically measured by some of the commer-
cially available OCTA devices (only Optovue has analytics available at this time)
with high accuracy and reproducibility. Attributable to their high resolution, these
objective metrics can be used to objectively evaluate the treatment response.

References
[1] Suzuki M, Sato T and Spaide R F 2014 Pseudodrusen subtypes as delineated by multimodal
imaging of the fundus Am. J. Ophthalmol. 157 1005–12
[2] Spaide R F and Curcio C A 2010 Drusen characterization with multimodal imaging Retina
30 1441–54
[3] Schmitz-Valckenberg S, Steinberg J S, Fleckenstein M, Visvalingam S, Brinkmann C K and
Holz F G 2010 Combined confocal scanning laser ophthalmoscopy and spectral-domain

8-13
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

optical coherence tomography imaging of reticular drusen associated with age-related

macular degeneration Ophthalmology 117 1169–76
[4] Zweifel S A, Spaide R F, Curcio C A, Malek G and Imamura Y 2010 Reticular
pseudodrusen are subretinal drusenoid deposits Ophthalmology 117 303–12
[5] Querques G, Canouï-Poitrine F and Coscas F et al 2012 Analysis of progression of reticular
pseudodrusen by spectral domain-optical coherence tomography Investig. Ophthalmol. Vis.
Sci. 53 1264–70
[6] Khan H A, Mehmood A and Khan Q A et al 2017 A major review of optical coherence
tomography angiography Expert Rev. Ophthalmol. 12 373–85
[7] Kashani A H, Chen C-L and Gahm J K et al 2017 Optical coherence tomography
angiography: a comprehensive review of current methods and clinical applications Prog.
Retin. Eye Res. 60 66–100
[8] Moult E M, Waheed N K and Novais E A et al 2016 Swept-source optical coherence
tomography angiography reveals choriocapillaris alterations in eyes with nascent geographic
atrophy and drusen-associated geographic atrophy Retina 36 S2–11
[9] Lauermann J L, Eter N and Alten F 2018 Optical coherence tomography angiography offers
new insights into choriocapillaris perfusion Ophthalmologica 239 74–84
[10] Sadda S R, Guymer R and Holz F G et al 2018 Consensus definition for atrophy associated
with age-related macular degeneration on OCT: classification of atrophy report 3
Ophthalmology 125 537–48
[11] Sacconi R, Borrelli E and Carnevali A et al 2020 Optical coherence tomography angiography
in type 3 neovascularization Diabetes and Fundus OCT Amsterdam: Elsevier pp 321–41
[12] Spaide R F, Jaffe G J and Sarraf D et al 2020 Consensus nomenclature for reporting
neovascular age-related macular degeneration data: consensus on neovascular age-related
macular degeneration nomenclature study group Ophthalmology 127 616–36
[13] Shi Y, Motulsky E H and Goldhardt R et al 2019 Predictive value of the OCT double-layer
sign for identifying subclinical neovascularization in age-related macular degeneration
Ophthalmol. Retin. 3 211–19
[14] Told R, Sacu S and Hecht A et al 2018 Comparison of SD-optical coherence tomography
angiography and indocyanine green angiography in type 1 and 2 neovascular age-related
macular degeneration Investig. Ophthalmol. Vis. Sci. 59 2393–400
[15] Zhao Z, Yang F and Gong Y et al 2019 The comparison of morphologic characteristics of
type 1 and type 2 choroidal neovascularization in eyes with neovascular age-related macular
degeneration using optical coherence tomography angiography Ophthalmologica. 242 178–86
[16] Nagiel A, Bansal M and Lenis T et al 2015 Morphological analysis of type 1, type 2, and
type 3 neovascularization in exudative age-related macular degeneration using OCT
angiography Investig. Ophthalmol. Vis. Sci. 56 3965–5
[17] Wang M, Zhou Y and Gao S S et al 2016 Evaluating polypoidal choroidal vasculopathy with
optical coherence tomography angiography Investig. Ophthalmol. Vis. Sci. 57 OCT526–32
[18] Tomiyasu T, Nozaki M, Yoshida M and Ogura Y 2016 Characteristics of polypoidal
choroidal vasculopathy evaluated by optical coherence tomography angiography Investig.
Ophthalmol. Vis. Sci. 57 OCT324–30
[19] Moult E, Choi W and Waheed N K et al 2014 Ultrahigh-speed swept-source OCT
angiography in exudative AMD Ophthalmic Surg. Lasers Imaging Retin. 45 496–505

8-14
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[20] El Ameen A, Cohen S Y and Semoun O et al 2015 Type 2 Neovascularization secondary to

age-related macular degeneration imaged by optical coherence tomography angiography
Retina 35 2212–18
[21] Souied E H, El Ameen A, Semoun O, Miere A, Querques G and Cohen S Y 2016 Optical
coherence tomography angiography of type 2 neovascularization in age-related macular
degeneration Dev. Ophthalmol. 56 52–6
[22] Tsai A S H, Cheung N and Gan A T L et al 2017 Retinal angiomatous proliferation Surv.
Ophthalmol. 62 462–92
[23] Kuehlewein L, Dansingani K K and de Carlo T E et al 2015 Optical coherence tomography
angiography of type 3 neovascularization secondary to age-related macular degeneration
Retina 35 2229–35
[24] Tan A C, Dansingani K K, Yannuzzi L A, Sarraf D and Freund K B 2017 Type 3
Neovascularization imaged with cross-sectional and en face optical coherence tomography
angiography Retina 37 234–46
[25] Miere A, Querques G, Semoun O, El Ameen A, Capuano V and Souied E H 2015 Optical
coherence tomography angiography in early type 3 neovascularization Retina 35 2236–41
[26] de Carlo T E, Romano A, Waheed N K and Duker J S 2015 A review of optical coherence
tomography angiography (OCTA) Int. J. Retin. Vitr. 1 5
[27] Wang R, Liang Z and Liu X 2019 Diagnostic accuracy of optical coherence tomography
angiography for choroidal neovascularization: a systematic review and meta-analysis BMC
Ophthalmol. 19 162
[28] Perrott-Reynolds R, Cann R and Cronbach N et al 2019 The diagnostic accuracy of OCT
angiography in naive and treated neovascular age-related macular degeneration: a review
Eye (Lond.) 33 274–82
[29] Schneider E W and Fowler S C 2018 Optical coherence tomography angiography in the
management of age-related macular degeneration Curr. Opin. Ophthalmol. 29 217–25
[30] Inoue M, Jung J J and Balaratnasingam C et al 2016 A comparison between optical
coherence tomography angiography and fluorescein angiography for the imaging of type 1
neovascularization Investig. Ophthalmol. Vis. Sci. 57 OCT314–23
[31] Faridi A, Jia Y and Gao S S et al 2017 Sensitivity and specificity of OCT angiography to
detect choroidal neovascularization Ophthalmol. Retin. 1 294–303
[32] Al-Sheikh M, Iafe N A, Phasukkijwatana N, Sadda S R and Sarraf D 2018 Biomarkers of
neovascular activity in age-related macular degeneration using optical coherence tomog-
raphy angiography Retina 38 220–30

8-15
 IOP Publishing

Photo Acoustic and Optical Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
Ayman El-Baz and Jasjit S Suri

Chapter 9
Optical coherence tomography angiography
imaging in age-related macular degeneration
Onur Gokmen and Muhammet Derda Ozer

Optical coherence tomography angiography (OCTA) is becoming a popular method

that enables non-invasive evaluation of retinal vascular structures. OCTA has
started to be used as an alternative method in age-related macular degeneration
(AMD) or in conjunction with classical fundus fluorescein angiography (FFA).
Although OCTA is a new method, it has not been able to replace FFA in AMD
completely. The limitations of OCTA for AMD can be listed as being highly
affected by image artifacts, providing a relatively small field of view, and the failure
of segmentation algorithms that can confuse the interpretation of the findings. In
this chapter, literature summaries about the usage areas and clinical application of
OCTA, especially in neovascular AMD, will be mentioned.

9.1 Introduction
AMD is a disease affecting the macula with progressive central vision loss. It is the
most common cause of irreversible vision loss in adults over 50 age in developed
countries and ranks third worldwide. Approximately 11 million people were affected
by this disease in the USA alone, and the prevalence of AMD in the world is 170
million [1]. Diabetes, hypertension, smoking, aging, ultraviolet exposure, obesity are
thought to be the risk factors in AMD development [2].

9.2 The principles Of OCTA technology and image quantification

methods
The most recent development in the OCT era is OCTA. We can obtain high quality,
3D angiographic images of the retinal and choroidal vascular layers through this
new technology. Despite its insensitivity to determining leakage and the other

doi:10.1088/978-0-7503-2060-3ch9 9-1 ª IOP Publishing Ltd 2021

 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

dynamic features of various retinal lesions seen in FFA, OCTA would likely help to
clarify the pathophysiological processes underlying multiple ocular diseases. The
repeated high-speed scans at the same location provide blood vessel identification.
The blood flow through vessels alters the reflectance signal, leading to variations
between the repeated scans at the same location. Quantifying these signal variations
makes it possible to identify vascular structures with an objective evaluation of
vascular pathologies [3]. 3D flow images can be presented as 2D en face images or
conventional cross-sectional images. The flow signal in the superficial (between the
internal limiting membrane and the outer limit of the outer plexiform layer) and the
deep (between the outer plexiform layer and Bruch membrane) retinal layers could
be combined and presented as one en face image in the originally mentioned way.
The colored flow rate could also be overlaid on the conventional OCT image in the
latter method. By these imagining techniques, the depth and topographical location
of abnormal vessels and blood flow impairment could be determined easily [4]. The
most commonly used OCTA parameters are vessel density and flow index [5–10].
The vessel density is defined as the percentage of the occupied area by the vessels,
while the flow index is the mean signal intensity of blood flow in a region of interest.
The diagnosis and follow-up of retinal vascular pathologies are possible with vessel
density analysis [11]. The flow index is more sensitive in determining metabolic
alterations in the retina [12].
There are various other biomarkers for quantification of vascular pathologies such
as avascular area, perfusion density mapping, neovascularization area that were
recently defined and utilized in monitoring the course of multiple diseases [13–18].
The avascular area measurement is a valuable tool for objectively assessing the
degree of capillary dropout in certain conditions, for instance, glaucoma and diabetic
retinopathy [13–15]. Perfusion mapping of vascular density is another biomarker to
evaluate flow impairment in a particular scanned area [17].
The measurement of neovascularization area, which is calculated as the total
neovascular area pixel, was defined recently to evaluate and monitor the retinal or
choroidal neovascular membranes quantitatively [18].

9.3 FFA, ICG, and OCTA

OCTA is a relatively new method that allows us to see retinal blood vessels’ flow by
creating high-resolution contrasts, non-invasively. In the b-scan scan in OCTA, the
images are obtained by observing the contrast created by erythrocytes’ flow in the
retinal vessels. For this reason, to create an image in OCTA, unlike many conven-
tional OCTs, it is necessary to obtain high-resolution images much faster [3].
FFA and ICG are conventional and invasive retinal vascular imaging methods,
and unlike OCTA, contrast dyes must be injected into the individual’s vessels in
these methods. Besides, FFA and ICG applications are more costly than OCTA.
The contrast agent should be dispersed in the vessels and retina for 10–30 min to
take images. This is a time-consuming practice with trained staff, so it is not ideal for
routine busy clinical practice. Another handicap of these methods is that even
though the dyes used in FFA and ICG are relatively safe, allergic reactions to these

9-2
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

contrast agents can still develop, nausea-vomiting and anaphylaxis can occur. For
these reasons, these methods should be applied in environments where healthcare
personnel can be treated urgently. Besides, since these drugs are excreted from the
body through the kidneys, patients should be questioned about renal failure, and
applications should be made by obtaining the relevant branch physician’s opinion
for patients with renal failure. In short, it would be appropriate to perform FFA and
ICG applications in clinics where emergency intervention opportunities are available
and where an internal medicine or nephrologist’s opinion can be obtained [19–21].
Although OCTA is a new imaging method, FFA is still considered as the gold
standard in wet AMD or retinal neovascularization. Regarding ICG, it is accepted
as the gold standard in imaging the choroidal vessels, especially since the contrast
specific to the choroidal vessels is used [22]. In these methods, images are formed in
two dimensions, but at the same time, unlike OCTA, images are obtained in a large
part of the retina. Unlike OCTA, in FFA and ICG, in diseases in which retinal
vascular integrity is impaired, the leaks created by the extravascular blood in the
vascular leakage areas and the regions where it is ponding can be analyzed.
However, since the images in FFA and ICG are two-dimensional, the desired
information about the lesion may not be obtained in any hemorrhage that will
develop in the retinal vessels, artifacts in the imaging of the vessels, opacities in the
cornea or lens, which are other parts of the eye, and defects that may affect the
image contrast below or above the retinal vessels. Also, changes in the retinal
architectural structure due to atrophy and fibrosis may make it challenging to
interpret OCTA and vascularity. As a result, FFA and ICG show the lesion’s
presence, but they cannot give us enough information about the lesion’s axial nature
[19, 21].

9.4 AMD types

Dry-type macular degeneration develops from the macula’s degeneration caused by
the accumulation of lipid deposits called drusen in the pigment epithelium with age.
In dry type AMD, the retinal pigment epithelium (RPE) pigmentary changes,
changes in the photoreceptor layers, and geographic atrophy are characterized by
widespread and progressive and incurable damage RPE, which is called geographic
atrophy, in the later stages of the disease can be seen [23, 24].
Wet AMD types are divided into 3. While neovascular vessels are seen in the sub-
retinal area in the first type (occult) CNV as type 1, neovascularization is observed in
the sub-RPE area in type 2 CNV (classical), and intra-retinal angiomatous
proliferation (rap) is observed in the third type. Thanks to OCTA’s 3D imaging
technology, it can be useful in differentiation in typing according to axial locations
in CNVs [25].

9.5 OCTA features in AMD

As mentioned earlier, the AMD is divided into two types during its natural course.
The dry type or non-exudative AMD typically affects the choriocapillaris and retinal
pigment epithelium layers leading to shrinkage of choriocapillaris and accumulation

9-3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

of drusenoid materials along with RPE alterations/atrophy. The wet type or

exudative AMD is related to the neovascularization and is classified into three
categories according to the neovascularization location as the occult (under the
RPE), the classical (under the retinal layers), and the retinal angiomatous prolifer-
ation type. The OCTA imaging has some limitations and superiorities to classical
imaging methods (FFA, ICG) in detecting certain features of both dry and wet
AMD.
OCTA is a fast, non-invasive imaging technique. It provides high resolution, 3D
static volumetric angiographic information from both superficial, intermediate, and
deep retinal plexus as well as choriocapillaris. On the other hand, the field of view is
limited in OCTA. It could not provide dynamic angiographic information and is very
sensitive to the patient’s fixation ability and ocular movements compared to the
conventional imaging methods. OCTA can provide important information with
regard to the diagnosis and treatment of AMD. Its 3D and high-quality image
acquisition features are unique for locating the neovascular membrane’s axial local-
ization and identifying its distinctive patterns. The static image acquisition feature also
provides a follow-up to the alterations that emerged after treatment [26–29]. On the
other hand, due to the lack of dynamic imaging features of OCTA compared to FFA,
it is challenging to evaluate the extent and severity of the neovascular lesion’s leakage
status [30]. Besides, the low flow rate cannot be distinguished from flow interruption
due to undetectable slow flow rate with interscan timing [31]. The scanned area is
relatively small compared to conventional imaging methods, and when one tries to
enlarge the scanning area, the image resolution would be significantly decreased due to
usage of insufficient A-scans to image a larger cross-sectional area [32–34].

9.6 Dry-type age-related macular degeneration

Dry-type AMD is characterized by drusen, pigmentary alterations, and RPE loss.
The role of vascular changes before the development of the alterations, as mentioned
earlier, is still controversial, but several histologic studies showed that choriocapil-
laris loss would have an initiator role in the development of dry-type AMD [35, 36].
Choriocapillaris layer visualization is limited by conventional angiography due to
poor visualization. Spectral-domain OCT devices have an inadequate role in
obtaining choriocapillaris layer images until a significant degree of RPE loss would
occur due to low short-wavelength beam penetration through the intact RPE [30].
After the invention of OCTA, the role of choriocapillaris perfusion in the develop-
ment of dry-type AMD was understood better. Lane et al showed a significant
decrease in choriocapillaris flow rate at the drusen site [37]. The generalized decrease
in choroidal vessel density that was mainly affecting the choriocapillaris layer has
previously been shown, and impairment of blood flow in choriocapillaris is
suspected as a crucial step in the pathogenesis of dry-type AMD [38].
The investigators preferred to investigate choroidal vascular alterations in geo-
graphic atrophy to avoid visualization difficulties that emerged by signal attenuation
due to intact RPE. Choi et al compared 12 eyes of seven patients with geographic
atrophy with control subjects using OCTA and found a significant flow impairment

9-4
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 9.1. Fundus photographs, OCT images and OCTA images of both eyes of a patient with dry-type
AMD.

in the choriocapillaris layer under the geographic atrophy area [39]. Besides, recently
published data shows a subtle flow impairment and diminished vessel density in the
choriocapillaris layer located in the vicinity of geographic atrophy. This implies that
the vascular changes would likely emerge before the development of RPE alterations
in dry AMD [40, 41]. OCTA also has superior efficacy in determining geographic
atrophy progression and excluding the presence of neovascularization by providing
additional blood flow parameters [42] (figure 9.1).

9.7 Wet-type age-related macular degeneration

The characteristic feature of wet-type or exudative AMD is neovascularization. It is
also further divided into three categories according to the axial location of the
choroidal neovascular membrane (CNVM) as occult, classical, and retinal angiom-
atous proliferation.
As mentioned previously, FFA as a gold standard imaging modality for detecting
wet-type AMD has not been replaced by OCTA yet. A series of studies evaluating
the sensitivity and specificity of OCTA in determining choroidal neovascularization
showed significantly high false-negative rates [43–47]. In other words, a relatively
large amount of choroidal neovascularisation that can be identified by FFA could
not be determined by OCTA. Based on the previous reports, CNVMs accompanied
by extensive subretinal hemorrhage and pigment epithelial detachment are more
prone to be missed by OCTA [43, 45, 46, 48]. In summary, type 1 (occult) wet-type
AMD has a lower detection rate on OCTA [47, 48]. The low detection rate of occult
CNVM by spectral-domain OCTA devices was related to the short wavelength laser
beam’s decreased penetration rate. Swept-source OCTA devices in which a longer
wavelength laser beam source is used may provide better visualization in such
conditions due to improved resolution [49]. In comparative studies, the same CNVM

9-5
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

was measured by both spectral-domain, and swept-source OCTA and the actual
CNVM dimensions were found to be larger with the latter one [50, 51]. A previous
study also showed that the measured CNVM dimensions were underestimated by
spectral-domain OCTA relative to ICG [52].
The reliability and effectivity of OCTA in determining CNV and measuring
actual lesion size are well studied in the literature. The studies subsequently
focused on the morphologic analysis of CNVM related to AMD in order to predict
activity and prognosis. The CNVMs related to the AMD were classified into two
distinct microvascular appearance patterns in further investigations as type 1 and 2
CNVM [27, 53–55]. Type 1 CNVM composed of densely packed vascular nets
showing a medusa pattern with extensive anastomosis. Type 2 CNVM has larger
and lower branching vascular networks that radiate from the main trunk with a

Figure 9.2. Fundus photograph of a patient with inactivate wet-type amd, subretinal druzenoid accumulation
in OCT and increased vascularity in deep capillary plexus in OCTA are observed.

9-6
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 9.3. Fundus photograph, FFA image, OCT image and neovascularization in deep vascular plexus in
OCTA of a patient with active wet AMD.

sharp demarcation and lower anastomosis rate [56–59]. Type 1 CNVM was found
to be associated with the activity [57–59]. Another activity sign was reported as the
presence of a perilesional, hypointense halo surrounding CNVM [57, 59]. Besides,
OCTA can also provide a quantitative assessment of CNVM membrane area and
its flow index [60]. These defining features of OCTA facilitate the follow-up of
CNVMs related to AMD and decision making on treatment initiation.
The alterations in vascular morphology of AMD related CNVM after anti-
VEGF treatment was reported in longitudinal studies [29, 61]. The short term
studies showed that anti-VEGF treatment leads to decreased CNVM flow area with
closure of anastomosis starting the first day after the treatment [29, 61, 62]. Besides,
it is shown that the CNVM begins to get activated after the second week of
treatment, which gives us an insight into the need for frequent treatment regimens
[61]. A recently published long term longitudinal study suggested that the anti-
VEGF treatment predominantly serves as anti-leakage rather than leading to
complete resolution of neovascular membranes [63] (figures 9.2 and 9.3).

9.8 Conclusion
As promising imaging technology, OCTA has made it possible to promptly diagnose
AMD related CNVM development and distinguish precisely its cross-sectional
location throughout the entire retina. Additionally, the quantification of the vessel
flow parameters and the definition of distinct features of active and inactive CNVM
membrane types by OCTA has facilitated the clinician’s capability to notice the
progression rates and earlier decision making on reinjection before any further visual
deterioration. With the lack of dye usage and the absence of FFA dynamic features
(leakage, pooling, etc), the imaging of obscured CNVMs is now possible. On the
other hand, there are multiple limitations in the visualization of type 1 CNVM due
to low penetration of scanning light under the RPE or the relatively small size of the
scanned area compared to conventional imaging methods. These limitations would

9-7
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

likely be overcome with the help of technological developments and the invention of
more suitable imaging methods like swept-source OCT.
The evolution of our understanding of AMD related CNVM, and geographic
atrophy progression has been enhanced after the clinical application of OCTA but is
still to be continued. The reliability and consistency of this recent imaging
technology on determining the progression of dry-type AMD with its unique
quantification features have been proved. Additionally, the types of CNVMs and
their response to anti-VEGF treatments could be visualized, and the reactivation
signs became easily noticeable imminently. In the future, up to date knowledge
would likely be influenced by further improvements in OCTA technology, and
proactive treatment options will be discussed in the future.

References
[1] Pennington K L and DeAngelis M M 2016 Epidemiology of age-related macular degener-
ation (AMD): associations with cardiovascular disease phenotypes and lipid factors Eye Vis.
3 34
[2] Lipecz A, Miller L and Kovacs I et al 2019 Microvascular contributions to age-related
macular degeneration (AMD): from mechanisms of choriocapillaris aging to novel inter-
ventions GeroScience 41 813–45
[3] Jia Y, Tan O and Tokayer J et al 2012 Split-spectrum amplitude-decorrelation angiography
with optical coherence tomography Opt. Express 20 4710
[4] Zhang M, Wang J and Pechauer A D et al 2015 Advanced image processing for optical
coherence tomographic angiography of macular diseases Biomed. Opt. Express 6 4661
[5] Jia Y, Bailey S T and Hwang T S et al 2015 Quantitative optical coherence tomography
angiography of vascular abnormalities in the living human eye Proc. Natl. Acad. Sci. 112
E2395–402
[6] Jia Y, Morrison J C and Tokayer J et al 2012 Quantitative OCT angiography of optic nerve
head blood flow Biomed. Opt. Express 3 3127
[7] Jia Y, Bailey S T and Wilson D J et al 2014 Quantitative optical coherence tomography
angiography of choroidal neovascularization in age-related macular degeneration
Ophthalmology 121 1435–44
[8] Wang X, Jia Y and Spain R et al 2014 Optical coherence tomography angiography of optic
nerve head and parafovea in multiple sclerosis Br. J. Ophthalmol 98 1368–73
[9] Jia Y, Wei E and Wang X et al 2014 Optical coherence tomography angiography of optic
disc perfusion in glaucoma Ophthalmology 121 1322–32
[10] Wei E, Jia Y and Tan O et al 2013 Parafoveal retinal vascular response to pattern visual
stimulation assessed with OCT angiography; S G Solomon PLoS One 8 e81343
[11] Al-Sheikh M, Tepelus T C, Nazikyan T and Sadda S R 2017 Repeatability of automated
vessel density measurements using optical coherence tomography angiography Br. J.
Ophthalmol 101 449–52
[12] Pechauer A D, Jia Y, Liu L, Gao S S, Jiang C and Huang D 2015 Optical coherence
tomography angiography of peripapillary retinal blood flow response to hyperoxia Investig.
Opthalmol. Vis. Sci. 56 3287
[13] Hwang T S, Gao S S and Liu L et al 2016 Automated quantification of capillary
nonperfusion using optical coherence tomography angiography in diabetic retinopathy
JAMA Ophthalmol 134 367

9-8
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[14] Liu L, Jia Y and Takusagawa H L et al 2015 Optical coherence tomography angiography of
the peripapillary retina in glaucoma JAMA Ophthalmol 133 1045
[15] Zhang M, Hwang T S, Dongye C, Wilson D J, Huang D and Jia Y 2016 Automated
quantification of nonperfusion in three retinal plexuses using projection-resolved optical
coherence tomography angiography in diabetic retinopathy Investig. Opthalmol. Vis. Sci.
57 5101
[16] Jain N, Jia Y and Gao S S et al 2016 Optical coherence tomography angiography in
choroideremia JAMA Ophthalmol 134 697
[17] Agemy S A, Scripsema N K and Shah C M et al 2015 Retinal vascular perfusion density
mapping using optical coherence tomography angiography in normals and diabetic
retinopathy patients Retina 35 2353–63
[18] Liu L, Gao S S, Bailey S T, Huang D, Li D and Jia Y 2015 Automated choroidal
neovascularization detection algorithm for optical coherence tomography angiography
Biomed. Opt. Express 6 3564
[19] Bennett T J, Quillen D A and Coronica R Fundamentals of fluorescein angiography Insight
41 5–11 http://ncbi.nlm.nih.gov/pubmed/30230734
[20] Kornblau I S and El-Annan J F 2019 Adverse reactions to fluorescein angiography: A
comprehensive review of the literature Surv. Ophthalmol 64 679–93
[21] Slakter J S, Yannuzzi L A, Guyer D R, Sorenson J A and Orlock D A 1995 Indocyanine-
green angiography Curr. Opin. Ophthalmol 6 25–32
[22] Desmettre T, Devoisselle J and Mordon S 2000 Fluorescence properties and metabolic
features of indocyanine green (ICG) as related to angiography Surv. Ophthalmol 45 15–27
[23] Kuppermann B and Narayanan R 2017 Hot topics in dry AMD Curr. Pharm. Des. 23 542–6
[24] Johnen S and Koutsonas A 2019 Trockene AMD—Zelluläre und gentherapeutische
Behandlungsansätze Klin. Monbl. Augenheilkd. 236 1096–102
[25] de Carlo T E, Romano A, Waheed N K and Duker J S 2015 A review of optical coherence
tomography angiography (OCTA) Int. J. Retin. Vitr. 1 5
[26] Spaide R F, Klancnik J M and Cooney M J 2015 Retinal vascular layers imaged by
fluorescein angiography and optical coherence tomography angiography JAMA Ophthalmol
133 45
[27] Farecki M-L, Gutfleisch M and Faatz H et al 2017 Characteristics of type 1 and 2 CNV in
exudative AMD in OCT-angiography Graefe’s Arch. Clin. Exp. Ophthalmol. 255 913–21
[28] Karacorlu M, Sayman Muslubas I, Arf S, Hocaoglu M and Ersoz M G 2019 Membrane
patterns in eyes with choroidal neovascularization on optical coherence tomography
angiography Eye 33 1280–9
[29] Lumbroso B, Rispoli M and Savastano M C 2015 Longitudinal optical coherence
tomography–angiography study of type 2 naive choroidal neovascularization early response
after treatment Retina 35 2242–51
[30] Schneider E W and Fowler S C 2018 Optical coherence tomography angiography in the
management of age-related macular degeneration Curr. Opin. Ophthalmol 29 217–25
[31] Couturier A, Mané V and Bonnin S et al 2015 Capillary plexus anomalies in diabetic
retinopathy on optical coherence tomography angiography Retina 35 2384–91
[32] Huang D, Jia Y, Gao S S, Lumbroso B and Rispoli M 2016 Optical coherence tomography
angiography using the optovue device Dev. Ophthalmol. 56 6–12
[33] Rosenfeld P J, Durbin M K and Roisman L et al 2016 ZEISS AngioplexTM spectral domain
optical coherence tomography angiography: technical aspects Dev. Ophthalmol. 56 18–29

9-9
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[34] Mase T, Ishibazawa A, Nagaoka T, Yokota H and Yoshida A 2016 Radial peripapillary
capillary network visualized using wide-field montage optical coherence tomography
angiography Investig. Opthalmol. Vis. Sci. 57 OCT504
[35] Biesemeier A, Taubitz T, Julien S, Yoeruek E and Schraermeyer U 2014 Choriocapillaris
breakdown precedes retinal degeneration in age-related macular degeneration Neurobiol.
Aging 35 2562–73
[36] Seddon J M, McLeod D S and Bhutto I A et al 2016 Histopathological insights into
choroidal vascular loss in clinically documented cases of age-related macular degeneration
JAMA Ophthalmol 134 1272
[37] Lane M, Moult E M and Novais E A et al 2016 Visualizing the choriocapillaris under
drusen: comparing 1050-nm swept-source versus 840-nm spectral-domain optical coherence
tomography angiography Investig. Opthalmol. Vis. Sci. 57 OCT585
[38] Cicinelli M V, Rabiolo A and Marchese A et al 2017 Choroid morphometric analysis in non-
neovascular age-related macular degeneration by means of optical coherence tomography
angiography Br. J. Ophthalmol 101 1193–200
[39] Choi W, Moult E M and Waheed N K et al 2015 Ultrahigh-speed, swept-source optical
coherence tomography angiography in nonexudative age-related macular degeneration with
Geographic Atrophy Ophthalmology 122 2532–44
[40] Sacconi R, Corbelli E, Carnevali A, Querques L, Bandello F and Querques G 2018 Optical
coherence tomography angiography in geographic atrophy Retina 38 2350–5
[41] Nassisi M, Shi Y and Fan W et al 2019 Choriocapillaris impairment around the atrophic
lesions in patients with geographic atrophy: a swept-source optical coherence tomography
angiography study Br. J. Ophthalmol 103 911–7
[42] Corbelli E, Sacconi R and Rabiolo A et al 2017 Optical coherence tomography angiography
in the evaluation of geographic atrophy area extension Investig. Opthalmol. Vis. Sci. 58 5201
[43] de Carlo T E, Bonini Filho M A and Chin A T et al 2015 Spectral-domain optical coherence
tomography angiography of choroidal neovascularization Ophthalmology 122 1228–38
[44] Gong J, Yu S, Gong Y, Wang F and Sun X 2016 The diagnostic accuracy of optical
coherence tomography angiography for neovascular age-related macular degeneration: a
comparison with fundus fluorescein angiography J. Ophthalmol. 2016 1–8
[45] Inoue M, Jung J J and Balaratnasingam C et al 2016 A comparison between optical
coherence tomography angiography and fluorescein angiography for the imaging of type 1
neovascularization Investig. Opthalmol. Vis. Sci. 57 OCT314
[46] Faridi A, Jia Y and Gao S S et al 2017 Sensitivity and specificity of OCT angiography to
detect choroidal neovascularization Ophthalmol. Retin. 1 294–303
[47] Malamos P, Tsolkas G and Kanakis M et al 2017 OCT-angiography for monitoring and
managing neovascular age-related macular degeneration Curr. Eye Res 42 1689–97
[48] Ahmed D, Stattin M and Graf A et al 2018 Detection of treatment-naive choroidal
neovascularization in age-related macular degeneration by swept source optical coherence
tomography angiography Retina 38 2143–9
[49] Považay B, Hermann B and Unterhuber A et al 2007 Three-dimensional optical coherence
tomography at 1050 nm versus 800 nm in retinal pathologies: enhanced performance and
choroidal penetration in cataract patients J. Biomed. Opt. 12 041211
[50] Novais E A, Adhi M and Moult E M et al 2016 Choroidal neovascularization analyzed on
ultrahigh-speed swept-source optical coherence tomography angiography compared to
spectral-domain optical coherence tomography angiography Am. J. Ophthalmol. 164 80–8

9-10
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[51] Miller A R, Roisman L and Zhang Q et al 2017 Comparison between spectral-domain and
swept-source optical coherence tomography angiographic imaging of choroidal neovascula-
rization Investig. Opthalmol. Vis. Sci. 58 1499
[52] Eandi C M, Ciardella A and Parravano M et al 2017 Indocyanine green angiography and
optical coherence tomography angiography of choroidal neovascularization in age-related
macular degeneration Investig. Opthalmol. Vis. Sci. 58 3690
[53] Dansingani K K and Freund K B 2015 Optical coherence tomography angiography reveals
mature, tangled vascular networks in eyes with neovascular age-related macular degener-
ation showing resistance to geographic atrophy Ophthalmic Sur. Lasers Imaging Retin 46
907–12
[54] Kuehlewein L, Dansingani K K and de Carlo T E et al 2015 Optical coherence tomography
angiography of type 3 neovascularization secondary to age-related macular degeneration
Retina 35 2229–35
[55] Kuehlewein L, Sadda S R and Sarraf D 2015 OCT angiography and sequential quantitative
analysis of type 2 neovascularization after ranibizumab therapy Eye 29 932–5
[56] Sulzbacher F, Pollreisz A, Kaider A, Kickinger S, Sacu S and Schmidt-Erfurth U 2017
Identification and clinical role of choroidal neovascularization characteristics based on
optical coherence tomography angiography Acta Ophthalmol 95 414–20
[57] Coscas G J, Lupidi M, Coscas F, Cagini C and Souied E H 2015 Optical coherence
tomography angiography versus traditional multimodal imaging in assessing the activity of
exudative age-related macular degeneration Retina 35 2219–28
[58] Al-Sheikh M, Iafe N A, Phasukkijwatana N, Sadda S R and Sarraf D 2018 Biomarkers of
neovascular activity in age-related macular degeneration using optical coherence tomog-
raphy angiography Retina 38 220–30
[59] Miere A, Butori P and Cohen S Y et al 2019 Vascular remodeling of choroidal neo-
vascularization after anti-vascular endothelial growth factor therapy visualized on optical
coherence tomography angiography Retina 39 548–57
[60] Hagag A, Gao S, Jia Y and Huang D 2017 Optical coherence tomography angiography:
Technical principles and clinical applications in ophthalmology Taiwan J. Ophthalmol 7 115
[61] Huang D, Jia Y, Rispoli M, Tan O and Lumbroso B 2015 Optical coherence tomography
angiography of time course of choroidal neovascularization in response to anti-angiogenic
treatment Retina 35 2260–4
[62] Pilotto E, Frizziero L and Daniele A R et al 2019 Early OCT angiography changes of type 1
CNV in exudative AMD treated with anti-VEGF Br. J. Ophthalmol 103 67–71
[63] Levine E S, Custo Greig E and Mendonça L S M et al 2020 The long-term effects of anti-
vascular endothelial growth factor therapy on the optical coherence tomography angio-
graphic appearance of neovascularization in age-related macular degeneration Int. J. Retin.
Vitr 6 39

9-11
 IOP Publishing

Photo Acoustic and Optical Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
Ayman El-Baz and Jasjit S Suri

Chapter 10
Optical coherence tomography angiography in
ophthalmology: an application in vessel imaging
Anadi Khatri, Araniko Pandey, Gunjan Prasai, Kinsuk Singh, Muna Kharel,
Eli Pradhan and Rupesh Agrawal

The posterior segment of the eye is an interface for a highly specialized neural
network that sends optical information to the brain. It is also a window to a wide
range of local and systemic pathophysiological processes. Employing clinical fundus
examination, various imaging technologies and biopsy samples, the retina can be
studied in diverse settings.
Because accurate diagnosis is a must for successful timely treatment, the identi-
fication can be notably improved by employing disease image modalities such as:
optical coherence tomography (OCT), fundus imaging and optical coherence tomog-
raphy angiography (OCTA). This chapter will focus on OCTA imaging for the
diagnosis of retinal diseases.
OCTA is an add-on to OCT that depicts information on retinal and choroidal
circulations without the need for invasive procedures such as fluorescein or
indocyanine dyes. With the development of OCT devices with high scan rates and
further improvements being made to adjust algorithms to make it optimal, OCTA
has increasingly become a powerful everyday non-invasive diagnostic tool in
ophthalmology.

10.1 Introduction
The posterior segment of the eye is an interface for a highly specialized neural network
that sends optical information to the brain [1, 2]. It is also a window to a wide range of
local and systemic pathophysiological processes. The retina is a directly visible
structure, which is a unique characteristic that allows extensive diagnostic oppor-
tunities [3]. Employing clinical fundus examination, various imaging technologies and
biopsy samples, the retina can be studied in diverse settings [4, 5].

doi:10.1088/978-0-7503-2060-3ch10 10-1 ª IOP Publishing Ltd 2021

 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Pathologies affecting the posterior segment of the eye are one of the major causes
of blindness in developed countries and are becoming more prevalent due to an
increasing population with greater longevity of life. Commonly encountered diseases
in the posterior segment worldwide are; uveitis, diabetic retinopathy, macular
edema, proliferative retinopathy, age related macular degeneration and glaucoma,
among others [6–11]. In recent times, studies from low- and middle-income countries
have also shown an increasing trend of retinal pathologies [9, 12].
Because accurate diagnosis is a must for successful timely treatment, the
identification can be notably improved by employing disease-specific computer-
aided diagnostic (CAD) systems based on different image modalities, such as OCT,
fundus imaging and OCTA. This chapter will focus on OCTA imaging for the
diagnosis of retinal diseases [13–17].
OCTA is an add-on to OCT that depicts information on retinal and choroidal
circulations without the need for invasive procedures such as the fluorescein or the
indocyanine dyes [17] (figure 10.1). Development of OCT devices with high scan
rates and further improvements being made to adjust algorithms to make it optimal,
OCTA has increasingly become a powerful everyday non-invasive diagnostic tool in
ophthalmology [18–20].
Since its invention in the early 1990s, OCT has become one of the most important
imaging modalities in ophthalmology [21]. OCT is a noniinvasive imaging technol-
ogy based on low-coherence interferometry. It generates high-resolution cross-
sectional images from backscattered light, enabling clinicians to assess structural
changes in different retinal diseases. However, structural OCT cannot be used to
monitor vascular changes because of the low contrast between capillaries and retinal
tissue [22, 23].
Given that many ocular diseases are associated with vascular abnormalities,
the ability to visualize and quantify blood flow in the eye is of high importance.
Conventionally, fluorescein angiography (FA) and indocyanine green angiography
(ICGA) are used for qualitative clinical assessment of retinal and choroidal
circulations, respectively [24]. They require intravenous injection of contrast agents,
which is time-consuming and can have potentially serious side effects [25]. Moreover,
FA and ICGA provide two-dimensional (2D) images of ocular circulations, limiting

Figure 10.1. OCTA (middle) showing all layers over the 9 mm × 9 mm scan over the posterior pole (inclusive
of both the optic nerve head and the macular area). Red free filter (left), infrared imaging (right).

10-2
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

depth perception and detailed investigation of the retinal and choroidal vasculatures.
The only limitation of OCTA over dye based angiography is the inability to identify
leakage [19, 22, 26, 27].
Multiple non-invasive imaging technologies have been employed in the last few
decades to visualize and quantify ocular circulations. Ultrasound color Doppler and
functional magnetic resonance imaging have been previously used for research
purposes [28]. While they were able to provide good tissue penetration, they did not
receive much clinical attention in ophthalmology due to poor resolution and
measurement reproducibility. Other imaging modalities such as laser Doppler
flowmetry and velocimetry, blue field entoptic technique and laser speckle assess-
ment have been used extensively to study vascular retinal physiology [29]. However,
their clinical use has been limited by their complexity, poor reproducibility and wide
population variation [30].
Functional extensions of OCT have also been explored for vascular imaging of
the eye. Doppler OCT (DOCT) uses the Doppler frequency shift resulting from the
movement of red blood cells to quantify volumetric blood flow in large vessels, as
well as for total retinal blood flow measurement [31, 32]. However, DOCT is less
suitable for investigating the retinal microvasculature. DOCT is sensitive to blood
flow parallel to the OCT beam, whereas flow in the retinal microvasculature is
mainly perpendicular to the OCT beam. OCT angiography (OCTA) is a more recent
development [10] and there are multiple terminologies used in interpretation which
are listed in table 10.1. It has the capability of producing high-resolution, 3D
angiograms of the retinal and choroidal vascular networks which provides more
detailed images of the vasculature.

10.2 Principles of OCTA

OCTA, a functional extension of OCT, generates high resolution volumetric en face
images from backscattered light. This allows clinicians and researchers to evaluate
retinochoroidal vasculature non-invasively. Recent advancements in the techniques
of image processing have proven the quality of scans comparable to or even better
than dye based procedures [19, 33].
OCTA images are generated by repeated scans of the retina at the same
location which detect the motion of RBCs within the vasculature against the static
background retinal tissues. Image processing was possible after the introduction
of Fourier-domain based OCT, which has a higher scanning speed [17, 34, 35]
(figure 10.2).
Various techniques were studied as a clinical tool for non-invasive angiography
viz. ultrasound color doppler, functional magnetic resonance imaging and doppler
OCT [32]. Poor reproducibility and complexity in the procedure made these
techniques less useful as a clinical alternative. Notably, the use of Doppler OCT
was deemed limited as it was less suitable for retinal microvasculature [36].
The variations in reflective signals of RBCs detected by the OCT can be
quantified on the basis of varying intensity, phase or complex (intensity and phase)
signals [27].

10-3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 10.1. Commonly used terminologies in OCTA.

Term Definition

A-scan rate The maximal rate at which A-scans are acquired during a B-scan.
Decorrelation The change in an area from one sequentially repeated image to another.
Gap defect An area where the image information is lost image which could be due to
various reason—such as eye movement, software rendering errors, etc.
Image artifact Abnormality in the visual representation of information derived from an
object (figure 10.9).
Phase The position of a wave cycle which describes or illustrates the difference
between periodic repetitions present in the wave.
Projection artifact Light which passes through the vessel fluctuate over time causing
anything posterior to the vessel to be illuminated by this fluctuating
light. This can be represented as artifact images of vessels and cause it
to be seen at deeper locations than they actually inhabit.
Segmentation Division of larger set into smaller sets according to a set of algorithms/
parameters.
Shadowing Attenuation of signal behind an obstruction which can be due to various
cause.
Stretch artifact A defect produced when the software attempts to correct eye motion,
causing a part of the image to be stretched.
Vessel doubling A defect produced when the software attempts to correct eye motion,
causing each blood vessel to be split and appear in parts or in the
whole mapped area (figure 10.6).
White line artifact White line seen in OCTA images associated with an eye movement.

Figure 10.2. Multiple A-scan lines which are combined to form the overall ‘surface’ image of the retina.

10-4
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

10.2.1 Variants
10.2.1.1 Intensity based OCTA
Uses amplitude or speckle information to generate an image. An intensity based
approach has been trialed in time-domain OCT to swept-source OCT systems. A
higher image quality and decreased signal-to-noise ratio (SNR) has been demon-
strated when any of the system operates with a faster scanning rate. Additional
computer algorithms have also been studied for better speckled variance informa-
tion between successive B-scans. These are correlation mapping and decorrelation
mapping—split spectrum amplitude decorrelation angiography (SSADA) [37].

10.2.1.2 Phase based OCTA

Uses the phase signals received from the tissues to generate an image. This was
basically used with Doppler OCT, where the Doppler shift of RBCs in motion was
recorded [38].

10.2.1.3 Complex signal based OCTA

Uses combined information of phase and amplitude-based variance and provides
superior results. The inability of amplitude-based variance to detect slow flowing
vessels and phase-based variance to detect microvasculature can be mitigated by a
combined approach. Recently, split-spectrum amplitude and phase gradient analysis
(SSAPGA) has been introduced, which combines phase variance with SSADA [37].
Complex signal based OCTA algorithms provide the best vascular visualization in
terms of connectivity, image contrast-to-noise ratio and SNR [39].

10.2.1.4 Metrics used in OCTA

Various metrics have been used for an objective quantification of vascular
pathologies. Vessel density and flow density of en face angiograms have been used
to clinically evaluate the retinal and choroidal vascular pathophysiological changes.
Quantifying avascular zones and neovascularization are other metrics that can be
used to monitor the progression of diseases like diabetic retinopathy [22, 27, 32].

10.2.2 Fundamentals of OCTA in ophthalmology

OCTA technology uses laser light reflectance of a moving column of red blood cells
(RBCs) to illustrate the vessels through different layers of the eye [22]. These layers can
be further divided into sub-layers depending on the tissue which is undergoing the
scanning. With current applications, this can be the retina, the choroid or even the neural
tissue such as the optic nerve head. The major advantage of this technique is that it is
non-invasive—meaning it eliminates the need for intravascular injection of dyes [33].
The OCT scan of a retina consists of multiple amplitude (A)-scans. This information
when translated into brightness (B)-scans via algorithms scan provide cross-sectional
structural information. The basic principal of OCTA is that it scans an area of the
same tissue area multiple times repeatedly and the differences are compared
between the scans, thus allowing the device to calculate and produce results [34].
These areas or zones are often divided into flow or no flow zones depending upon
the detectable thresholds of the device. Usually, the scan rates, at the time of writing

10-5
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

of this chapter, are limited to four scans per second on any segment of the retina
[40, 41]. We can highly expect there to be developments of machines with higher
scan rates or adjustable scan rates to detect such flows in future.

10.2.3 Scanning technique

Light, usually of wavelength near to 800 nm, or closer to 1050 nm, is emitted
depending upon the type of OCT. The former is usually used in spectral domain
OCT (SD-OCT) while the latter with a longer wavelength is generally used in swept-
source OCT (SS-OCT) [42]. The benefits of having longer wavelengths is that they
have deeper tissue penetrance and hence can provide information on deeper tissues.
Both the retinal and the choroidal microvasculature can be visualized with help of
OCTA. However, this same property leads to a disadvantage of generating a slightly
lower axial resolution of the scanned areas [43].
In contrast, if we opt for dye-based angiography, fluorescein angiography FA is
generally only applicable for evaluating the status of the retinal vessels and
indocyanine green is more ideal for imaging the choroid [44].
OCTA works on the principle of diffractive particle movement detection and
mainly uses two methods to detect motion: amplitude decorrelation and phase
variance. The former detects amplitude differences of two different B-scans. Phase
variance is more related to the properties of the light wave which is emitted. It
calculates the variation of phase when moving objects are detected [22, 26, 27].
By using these techniques and incorporating various motion algorithms, OCTA is
able to provide flow information. The only major limitation of OCTA is that it
cannot appreciate leakage and that is where dye based imaging still outshines it.
However, due to its ability to scan various layers and detect flow individually in each
of them, OCTA does have an advantage that it can provide approximate delineation
and size of retinal or choroidal pathology such as the choroidal neovascularization/
membrane [45] (figure 10.3). This is especially superior in detection and diagnosis of
type 1 CNV where localization is inferential and most of the time is very inaccurate

Figure 10.3. OCTA eliciting choroidal neovascularization in the external retinal layer extending from the
choriocapillaris. The image color is inversed for easier depiction.

10-6
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

with imaging that uses dyes. However, one must remember that OCTA scans can be
obscured by hemorrhage in any layer of the retina or the choroid as they block the
scanning light and can give an impression of flow voids [46] (figure 10.4).
In the present scenario, the main problem associated with OCTA imaging is that
it is more prone to artifacts than conventional dye-based angiography. The larger
retinal vessels produce a ‘ghost image’—often called the shadow artifact. This is
usually a problem when the outer retinal slabs are being evaluated as these artifacts
make it difficult to appreciate the presence of abnormal vessels in the deeper slabs.
At the same time, the light incident over the retinal vessels may be refracted or even
pass through, causing an increased illumination in the deeper layers. These are
termed projection artifacts [32, 40, 46] (figure 10.5).
As discussed above, OCTA uses the principle that movement represents blood
flow. This makes it prone to motion artifact also. White bands or lines (which denote
decorrelation signal over the entire B-scan) appear if the patient’s eye moves due to
fixation loss (figure 10.6 top left). Similarly, if the OCTA registers a signal as no
movement, there will be the presence of black bands or links. Usually when the
patient blinks, the OCT signal is blocked and is unable to detect any movement
(figures 10.6 top left and 10.7).

Figure 10.4. OCTA scans of the external retina and choriocapillaris obscured due to scarring which were present
in the inner retinal layers. Note the black artifact in the middle image giving a false impression of flow void.

Figure 10.5. Example of projection artifacts where the vessels from the superficial retinal plexuses are being
represented in the deeper retinal plexuses due to over illumination due to high flow signal detection which is
overexpressed.

10-7
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 10.6. Black and white line artifacts which are present in the top left and top right images. Black lines
are produced during the blinking phase where the signals are blocked and white lines are produced due to
saccades produced by eye movements, which OCTA expresses as a ‘particle movement’ and registers as a gross
flow. Also note the vessel doubling and stretch effects in the vessels due to the attempt by the algorithm to
correct and correlate the signals.

Figure 10.7. Example of black band artifact.

10-8
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

10.3 OCTA of normal eyes

Current OCTA devices can obtain volumetric scans at up to 100 000 A-scans per
second in approximately 3–4 s. The scan areas are often offered from an option of
2 × 2 mm, 3 × 3 mm, 6 × 6 mm and 9 × 9 mm OCT angiograms [22, 41, 46]
(figures 10.8 and 10.9).
The software automatically segments the scanned areas of the retinal thickness
into ‘superficial’ and ‘deep’ inner retinal vascular plexuses, outer retina or avascular
layer and choriocapillaris. While the OCT angiogram segmentation of the superficial
inner retina contains a projection of the vasculature in the retinal nerve fiber layer
(RNFL) and ganglion cell layer (GCL), the deeper inner retina slabs are represented
as a composite of the vascular plexuses at the border of the inner plexiform layer
(IPL) and inner nuclear layer (INL) and the border of the INL and outer plexiform
layer (OPL) [47].
The OCTA prototype with the fastest acquisition rate can obtain scans of 500 × 500
A-scans at 400 000 A-scans per second in less than 4 s. This ultra-high speed allows for
imaging of wider fields of view and uses 1060 nm wavelength which allows increased
light penetration into pigmented tissues and improved choroidal blood flow visual-
ization [40, 48].

10.4 Current applications of OCTA in ophthalmology

10.4.1 OCTA in dry (non-neovascular) AMD
Dry age-related macular degeneration (AMD) is characterized by the presence of
drusen, pigmentary changes in retina and loss of photoreceptors and RPE. It is also

Figure 10.8. OCT A-scan (represented as horizontal lines) and the scan is being represented from nasal to
temporal side. On the top, some spikes are present. This is usually due to eye movements or blinking that might
have occurred during the scanning of that region.

10-9
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 10.9. An image artifact causing the retinal slabs to be seen as separated in full thickness scan (similar to
tectonic plates). These are usually caused when some signal loss occurs during the scan in that area but the
OCT then tries to ‘stitch’ together the scans to create a complete image.

known as geographic atrophy (GA) of the macula. It is a well-established fact that

there is decreased foveolar choroidal blood flow in AMD and it has been
hypothesized that the evaluation of the choroidal vascularity may help to predict
the disease progression [49].
Various literatures have already pointed that there are areas of impaired flow in
choriocapillaris which typically extend beyond the delineation of the GA. Eyes with
dry AMD were also found to have a generalized decrease in choriocapillaris density
and often associated with presence of drusen. Other studies have further added that
these alterations were not due to shadowing or signal blockages.
Further studies in future can ascertain these findings to determine if the
choriocapillary changes associated with drusen are true areas of flow changes.

10.4.2 OCTA in wet (neovascular) AMD

With the introduction of OCTA, this novel technology has offered further insight
into our understanding of the disease pathology and has enabled us to monitor
choroidal neovascularization (CNV) in the majority of cases [49–51]. Various studies
have already shown that sensitivity and specificity for CNV detection with en face
OCTA combined with cross-sectional OCTA can be almost as effective as the gold

10-10
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

standard fluorescein angiography (FA) with OCT, hence avoiding the need of
invasive procedures (figure 10.3).
There have been several publications concerning OCTA of eyes with wet AMD.
The ability of OCTA to elicit CNV was first described in 2014 and since then various
literatures are present describing the sensitivity and specificity using FFA/ICG as the
gold standard for comparison [52]. Much work and effort have also been made to
understand and describe characteristics of CNV both quantitatively and qualita-
tively. It is beyond the scope of this chapter to describe the various types of lesions
such as type 2 or type 3 CNV (retinal angiomatous proliferation, RAP), and further
reading is recommended from literature available in the research database [53].
The major advantage of OCTA in monitoring the response of treatment or the
progression of the pathology is that the patient can undergo eye scan multiple times
in every follow-up without the need of dye injection. The eyes can also be analyzed
in segments or slabs to fully understand the extent of the disease and damage to the
internal retinal elements.

10.4.3 OCTA in diabetic retinopathy

Various researches have demonstrated the advantage and ability of OCTA to detect
and even predict the progression of diabetic retinopathy [54, 55]. Depending upon
the features such as choriocapillary abnormalities and/or retinal microvascular
abnormalities such as microaneurysms, vascular remodeling adjacent to the foveal
avascular zone (FAZ), enlarged FAZ, capillary tortuosity and dilation, OCTA has
even able to predict the likelihood of progression of diabetic retinopathy. Used in
various combinations, it may even predict which of the features can mean the eye is
more susceptible to becoming potentially blind [26, 27, 56–58].
Various sizes and shapes of microaneurysms have been analyzed. Microaneurysms
have also been classified as high flow and low flow depending upon the signal threshold
[59]. The location where they can cause more anatomical and functional damage has
been also postulated. Various literatures have already identified that if the micro-
aneurysms are present more on the deep capillary plexus, they cause the retina to be
more vulnerable to edema [26, 27, 58] (figure 10.10). Recent studies have shown a

Figure 10.10. Microaneurysm depicted by OCTA at various layers of the retina.

10-11
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

similar distribution of the low flow microaneurysms—meaning those which were able
to be caught in the FFA but not in OCTA have been found in areas where the capillary
density is decreased [26]. It has been postulated that these may be the areas of
decompensated flow and such microaneurysms could mean the areas of impending
ischemia and ischemic maculopathy in the long term [27].
Some studies have actually shown that OCTA may be superior in detection of
microaneurysms as it was able to appreciate some microaneurysms that were
not detected by conventional angiography [60]. OCTA has also been shown to
successfully detect other abnormalities which may not be evident on FA such as
areas of retinal non-perfusion, reduced capillary density and increased vessel
tortuosity.

10.4.4 OCTA in artery and vein occlusion

Retinal vascular occlusions are one of the most common retinal vascular pathologies
[61] which are slowly being understood more using OCTA as an imaging modality.
Various reports have suggested that OCTA usually shows large areas of capillary
non-perfusion (flow voids) in the area of ischemia with clear delineation of the
boundary, including vascular abnormalities such as microaneurysms, telangiectasia
and anastomoses [62, 63]. Vascular density mapping can demonstrate the decrease in
the density of vascular plexuses—which directly is also a representation of the
ischemic load. The same area can be measured using an inbuilt caliper or using the
disc diameters to estimate the ischemic stress of the retina [64].
Branched retinal artery occlusion demonstrates wedge-shaped areas of capillary
non-perfusion that correlates to areas of abnormalities on the retinal thickness map.
This illustrates the potential use of OCTA in pinpointing areas of ischemia and
edema (figure 10.11).

Figure 10.11. An OCTA image showing flow voids in the inferotemporal arcade following inferotemporal
tributary vein occlusion.

10-12
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Central retinal artery occlusions show diffuse capillary flow voids areas supplied
by the central retinal artery in both the superficial capillary plexus and the deep
capillary plexus [65]. Flow is still seen in the major retinal vessels. There is an
absence of blood flow in the superficial disc vasculature supplied by the central
retinal artery around the optic disc but the flow around the lamina cribosa remains
intact.
One major disadvantage is that OCTA provides a snapshot in time and registers
all the movements that are above its detection threshold. Hence, it is not able
demonstrate delayed arteriovenous transit time. However, the other major advant-
age is that OCTA also enables one to view the choriocapillaris and choroidal
vascular flow—which are usually found to be minimally affected in arterial or
venous occlusions [66, 67].

10.4.5 OCTA in glaucoma

OCTA has been a prolific tool for evaluating optic disc perfusion in eyes suspected
with glaucoma or already with signs of the disease. OCTA usually demonstrates the
dense peripapillary microvascular network in normal eyes and the current evidence
is that this vascular network is attenuated in both the superficial disc vasculature
and the deeper lamina cribosa (figure 10.12). With processing after averaging the
decorrelation signal in OCT angiograms, it is able to further approximate the area of
microvasculature and provide physicians with parameters allowing them to calculate
the flow index. This flow index has been depicted to have very high sensitivity and
specificity in differentiating glaucomatous eyes from normal eyes and there is
evidence that it may even help in segregating eyes at risk of glaucoma in future or
eyes with pre-perimetric glaucoma [14, 68].
The concept of assessing glaucomatous damage by evaluating blood flow is not
new. Over decades, physicians and researchers working for drug delivery and
understanding the pathogenesis of glaucoma have been correlating between blood
flow and glaucomatous damage to the optic nerve head. Fluorescein angiography
features such as fluorescence and delayed vascular filling have been noted. However,

Figure 10.12. OCTA scan showing vascular arrangement at different layers of the optic nerve head (top row).
Vascular density elicited by the OCTA using heat mapping colors (lower row middle).

10-13
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

this method is invasive and is not suitable for immediately repeating if the initial
results are inconclusive, making it impractical for routine evaluation to understand
the trend of the damage [69, 70].
OCTA has the ability to compare both dysfunctional and dead ganglion cells and
various hypotheses correlating metabolism demands of such neuronal structures
with vascular density and parameters like papillary, para or peripapillary vascular
density have established themselves as strong indicators of assessing glaucomatous
damage. There are even clinical findings suggesting that reduced capillary plexus
could be representing areas of RFL or GCC which are under stress and may
undergo apoptosis causing irreversible damage if there is no intervention [71].
Prospective cohorts of these eyes have shown that such areas undergo subsequent
thinning of NFL and GCC—further strengthening the association. One must
remember that although this thinning can be detected by structural OCT and one
can monitor the progression and modify treatment accordingly, irreversible damage
will have already occurred [72]. By adding OCTA as a diagnostic modality, it has an
advantage of detecting glaucoma earlier, perhaps even before the damage becomes
evident, enabling physicians to adopt treatment at primary prevention level.

10.4.6 OCTA in uveitis

10.4.6.1 Using OCTA for retinal imaging in uveitis
Currently, there are various researches which have pointed out the usefulness of
OCTA in evaluation of the disease progression or response of treatment of uveitis.
This has been limited not just for posterior uveitis but also for anterior uveitis [73].
Recent studies using OCTA have showed that uveitic macular edema is
associated with changes in the capillary density or morphology at deep capillary
plexus (DCP). Eyes with uveitis and macular edema were found to have a
significantly lower vessel density in DCP [15]. Parafoveal capillary loss in the
superficial capillary plexus was also observed irrespective of whether the macular
edema was present or not. These findings have been further reinforced with a similar
finding where remodeling of capillaries and the irregularity and/or enlargement of
foveal avascular zone (FAZ) were found to be consistent in patients with uveitis [74].
However, one must again not forget the basic principle of OCTA. It detects
movement of blood flow and there is a threshold for which the signal can be reseated
[22]. The areas of non-perfusion may only be representing sluggish or a slow flow
not definitely a complete capillary dropout/or inadequate capillary perfusion [75].
Irrespective of the pathology, it is also important to remember that most of the
vasculopathies related with retina and choroid have shown that the insult in the DCP
is of more serious nature [15, 26, 73]. One of the reasons is that the deep capillaries
may be more susceptible to ischemia as they are not directly connected to arterioles.
However, production of projection artifacts on OCTA which maybe produced by the
atrophic or dead retinal tissues must also be considered, which may confuse by
indicating nonperfusion/hypoperfusion in normal areas. If such confusions occur, a
rescan and careful en face revaluation by the physician or technician must be done to
assess each of the slabs in the machine and not to rely only on the printed report.

10-14
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

10.4.6.2 Using OCTA for choroidal imaging in uveitis

Evaluation of the choriocapillaris by OCTA has become an important indicator to
monitor disease progression. These have been demonstrated in various case reports
and literature [13, 76–78]. Choriocapillaris is noted to have loss in the ‘fine granular
pattern’ which it normally elicits and instead becomes irregular with loss of the
architecture [77] (figure 10.13). This may further be evaluated by measuring the
choroidal thickness. Involvement of the choroid in uveitis and evaluation of
thickness have already been established with development of a novel quantitative
imaging biomarker, called the choroidal vascularity index (CVI) [78].
The index is defined as the ratio of vascular area under scan to the total choroidal
area in reference. CVI has been indicated as a potential tool in establishing early
diagnoses, monitoring disease progression and prognosticating patients and has the
potential to be a robust marker in numerous retinal and choroidal diseases.
Uveitis is also associated with development of CNV in many of the uveitis
disorders such as punctuate inner choroiditis or in ocular histoplasmosis syndrome.
The challenge is, however, differentiating purely inflammatory lesions from
inflammatory CNV in uveitis patients [79].

Figure 10.13. Normal OCTA of the choriocapillaris showing granular angioarchitecture. Also note the
projection artifacts from the retinal vessels.

10-15
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

OCTA is known to aid early detection of CNV, especially when findings from
other imaging modalities become inconclusive. Similar to its application in wet-
ARMD, the same property allows it to delineate the neovascular network inside the
area of a lesion without being obscured by dye leakage [80]. OCTA is also effective
in both detecting and revealing CNV despite existing subretinal fluid or hemorrhage
around the area of the lesion which sometimes is obscured, prevailing in such cases
over FFA. Its effectiveness in monitoring the progress of CNV lesions that were
under treatment and their response to it is also remarkable. Lesions that show no
blood flow signal on OCTA can be distinguished from potential CNV lesions and
further management can be planned accordingly [81].
A short summary on application of OCTA and findings in each of the scanned
slabs in uveitis are listed below [15]:
• Superficial retinal capillary plexus—found to be disturbed in inflammatory
vasculitis. OCTA reveals flow voids in superficial retinal vessels. May also
show capillary remodeling and a lower vessel density. The vascular
angioarchitecture can be interpreted, which may be useful in management
decisions [74].
• Deep retinal capillary plexus—OCTA is able to detect patterns when such
eyes are associated with cystoid macular edema. The disease or pathology
may be more vision threatening [82].
• Choriocapillaris—there is decrease in the choroidal blood flow or ischemia.
These areas of flow voids are found to be very consistent with hypo-perfused
areas when the same are compared with the findings from the indocyanine-
green angiography findings [15].

10.5 The future of OCTA and ophthalmology

10.5.1 Non-invasive imaging and neurovascular coupling
Recently, studies of blood flow dynamics are being concentrated in disruption of
autoregulatory mechanisms of the retinal blood flow. Various human as well as
animal studies are being conducted to identify quantifiable parameters suggesting
neurovascular decoupling, in advance of pathologies like glaucoma and diabetic
retinopathy. Neurovascular coupling is the hemodynamic response to light stim-
ulation, which has been shown by high correlation of increased blood flow in the
retina and optic nerve head (ONH) to increased neural activity [83, 84].
With the advancement of newer technologies to identify the pattern of blood flow
and vascular pathologies in retina or ONH, it has become possible to explore the
implications of this physiologic phenomenon. Laser Doppler flowmetry is based on
the Doppler effect which uses a fundus camera and a computer system to calculate
velocity, blood volume and blood flow around the area of interest. OCTA can be
used to compute the flow index, which is an arbitrary measurement of blood flow.
Color Doppler imaging is also based on the principle of Doppler frequency shift and
can be used to measure the pattern of blood flow changes in color coded format.
Laser speckled flowgraphs use speckled contrast images of the retina and ONH to

10-16
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

present a temporal waveform pattern of blood flow and hence, provide values of
blood flow and velocity for analysis [83].
There are numerous trials that are studying on retinal neurovascular response to
visual stimulus. These experiments have presented multiple protocols, to establish a
physiological basis of the dose/stimulus-response relationship. The majority of the
studies used flicker stimulus in dark adapted and light adapted eyes within a certain
time frame. The neurovascular response to visual stimuli are then recorded using
various machines ranging from OCTA, laser speckled contrast imaging to scanning
laser ophthalmoscope. These experiments are slowly gaining momentum for trans-
lating this theory into the mainstream clinical practice [85–89].
Studies have shown changes in the compliance of retinal vessels with stimulation
of the ganglion cells to light. The normal response to visual stimuli that demands
high metabolism is vasodilation of the retinal blood vessels [83]. This physiologic
autoregulatory response can be disrupted early during the pathologic course of
diabetic retinopathy or glaucoma, which is termed as neurovascular decoupling. The
purpose of these ongoing experiments is to quantitatively identify and use these
biomarkers for early diagnosis and treatment [88]. The morphologic areas of interest
that can be quantified for correlation can be the retinal vasculature at various levels
viz. superficial, middle or deep capillary plexus around the parafoveal region, optic
nerve head or elsewhere. Change in blood flow can be quantitatively measured with
biomarkers that calculate the velocity, blood flow or vascular cross-sectional area
and give us values like blood flow velocity, vessel length or vessel length diameter.
These values can then be clinically correlated to predict diabetic retinopathy,
glaucoma or even response to treatment [90].
Blood vessels dilate to meet the metabolic demands of the retinal neurons in
response to visual stimuli [84, 85]. Preliminary studies have shown decreased retinal
vascularity in mild DR, in contrast to an increased vessel density in diabetics without
DR. These studies postulate varying patterns of decreased retinal vessel density or
increased foveal avascular zone at various stages of DR [88].
Similarly, disrupted autoregulatory neurovascular response around the optic
nerve head has been postulated in those with glaucoma. This can be of particular
interest for developing diagnostic biomarkers in advance of any visual field defect or
neuronal damage in glaucoma [83].
In conclusion, efforts have now been concentrated towards developing an easy
and replicable protocol with least variability. Pairing this to a cost-effective, non-
invasive and user-friendly imaging device with a robust computational analytic
system, will certainly narrow down our search for quantifiable biomarkers in the
management of ocular neurovascular pathologies.

References
[1] Hogan M J, Alvarado J A and Weddell J 1971 Histology of the Human Eye (Philadelphia,
PA: W.B. Saunders)
[2] Jakobiec F A 1982 Ocular Anatomy, Embryology and Teratology (Philadelphia, PA: Harper
and Row)

10-17
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[3] Hildebrand G D and Fielder A R 2011 Anatomy and physiology of the retina Pediatric
Retina (Berlin: Springer) pp 39–65
[4] Schneiderman H 1990 The funduscopic examination ed H K Walker, W D Hall and J W
Hurst Clinical Methods: The History, Physical, and Laboratory Examinations 3rd edn
(Boston, MA: Butterworth) ch 117. Available from https://ncbi.nlm.nih.gov/books/NBK221/
[5] Khochtali S, Khairallah-Ksiaa I and Ben Yahia S 2016 Normal fundus fluorescein
angiography Uveitis Atlas [Internet] ed V Gupta, Q D Nguyen, P LeHoang and C P
Herbort Jr (New Delhi: Springer) pp 1–7
[6] Wong W L, Su X, Li X, Cheung C M, Klein R, Cheng C Y and Wong T Y 2014 Global
prevalence of age-related macular degeneration and disease burden projection for 2020 and
2040: a systematic review and meta-analysis Lancet 2 E106–16
[7] Silva R, Cachulo M L, Fonseca P, Bernardes R, Nunes S, Vilhena N and Faria de Abreu J
2011 Age-related macular degeneration and risk factors for the development of choroidal
neovascularisation in the fellow eye: a 3-year follow-up study Ophthalmologica 226 110–8
Epub 2011 Aug 3
[8] Klein R, Klein B E, Moss S E, Davis M D and DeMets D L 1984 The Wisconsin
epidemiologic study of diabetic retinopathy. III. Prevalence and risk of diabetic retinopathy
when age at diagnosis is 30 or more years Arch. Ophthalmol. 102 527–32
[9] Friedman D S, Ali F and Kourgialis N 2011 Diabetic retinopathy in the developing world:
how to approach identifying and treating underserved populations Am. J. Ophthalmol. 151
192–4 e1
[10] Tham Y-C, Li X, Wong T Y, Quigley H A, Aung T and Cheng C-Y 2014 Global prevalence
of glaucoma and projections of glaucoma burden through 2040: a systematic review and
meta-analysis Ophthalmology 121 2081–90
[11] Robert N and Scott W 2010 Uveitis: Fundamentals and Clinical Practice 4th edn (Maryland
Heights, MO: Mosby Elsevier) pp 303–18
[12] Hubley J and Gilbert C 2006 Eye health promotion and the prevention of blindness in
developing countries: critical issues The Br. J. Ophthalmol. 90 279–84
[13] Agrawal R, Ding J, Sen P, Rousselot A, Chan A and Nivison-Smith L et al 2020 Exploring
choroidal angioarchitecture in health and disease using choroidal vascularity index Prog.
Retin. Eye Res 77 100829
[14] Jia Y, Morrison J C and Tokayer J et al 2012 Quantitative OCT angiography of optic nerve
head blood flow Biomed. Opt. Express 3 3127–37
[15] Pichi F, Sarraf D and Arepalli S et al 2017 The application of optical coherence tomography
angiography in uveitis and inflammatory eye diseases Prog. Retin. Eye Res 59 178–201
[16] Koustenis A, Harris A, Gross J, Januleviciene I, Shah A and Siesky B 2017 Optical
coherence tomography angiography: an overview of the technology and an assessment of
applications for clinical research Br. J. Ophthalmol 101 16–20
[17] Spaide R F, Klancnik J M and Cooney M J 2015 Retinal vascular layers imaged by
fluorescein angiography and optical coherence tomography angiography JAMA Ophthalmol
133 45–50
[18] Choi W, Mohler K J, Potsaid B, Lu C D, Liu J J and Jayaraman V et al 2013
Choriocapillaris and choroidal microvasculature imaging with ultrahigh speed OCT angiog-
raphy PLoS One 8 e81499
[19] de Carlo T E, Romano A, Waheed N K and Duker J S 2015 A review of optical coherence
tomography angiography (OCTA) Int. J. Retin. Vitr 1 5

10-18
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[20] Matsunaga D, Puliafito C A and Kashani A H 2014 OCT angiography in healthy human
subjects Ophthalmic Surg. Lasers Imaging Retin 45 510–5
[21] Gabriele M L, Wollstein G and Ishikawa H et al 2011 Optical coherence tomography:
history, current status, and laboratory work Invest. Ophthalmol. Vis. Sci. 52 2425–36
[22] Huang D, Swanson E A and Lin C P et al 1991 Optical coherence tomography Science 254
1178–81
[23] Fercher A F, Hitzenberger C K, Drexler W, Kamp G and Sattmann H 1993 In vivo optical
coherence tomography Am. J. Ophthalmol. 116 113–4
[24] Schachat A P Clinical applications of diagnostic indocyanine green angiography Ryan’s
Retina vol 1 6th edn (London: Elsevier) pp 46–76
[25] Hope-Ross M, Yannuzzi L A, Gragoudas E S, Guyer D R, Slakter J S and Sorenson J A et
al 1994 Adverse reactions due to indocyanine green Ophthalmology 101 529–33
[26] Khatri A, Pradhan E, Bal Kumar KC, Kharel M, Rijal R K and Timalsena S et al 2020
Detection, localization, and characterization of vision-threatening features of microaneur-
ysms using optical coherence tomography angiography in diabetic maculopathy Eur. J.
Ophthalmol 31 1120672120924609
[27] Khatri A, Bal Kumar K C, Kharel M, Ashma K C and Pradhan E 2020 Analysis of
microaneurysms and capillary density quantified by OCT-angiography and its relation to
macular edema and macular ischemia in diabetic maculopathy Eye [Internet] 35 1777–9
[28] Lieb W E, Cohen S M, Merton D A, Shields J A, Mitchell D G and Goldberg B B 1991
Color Doppler imaging of the eye and orbit Arch. Ophthalmol. 109 527–31
[29] Neganova A Y, Postnov D D, Jacobsen J C and Sosnovtseva O 2016 Laser speckle analysis
of retinal vascular dynamics Biomed. Opt. Express 7 1375–84 Published 2016 Mar 18
[30] Basak K, Manjunatha M and Dutta P K 2012 Review of laser speckle-based analysis in
medical imaging Med. Biol. Eng. Comput. 50 547–58
[31] Tan O, Liu G, Liang L, Gao S S, Pechauer A D and Jia Y et al 2015 En face Doppler total
retinal blood flow measurement with 70 kHz spectral optical coherence tomography J.
Biomed. Opt. 20 066004
[32] Wang Y, Fawzi A, Tan O, Gil-Flamer J and Huang D 2009 Retinal blood flow detection in
diabetic patients by Doppler Fourier domain optical coherence tomography Opt. Express 17
4061–73
[33] Kwiterovitch K A, Maguire M G and Murphy R P et al 1991 Frequency of adverse systemic
reactions occurring after fluorescein angiography—results of a prospective study
Ophthalmology 98 1139–42
[34] Wylęgała A 2018 Principles of OCTA and applications in clinical neurology Curr. Neurol.
Neurosci. Rep. 18 96 Published 2018 Oct 18
[35] Spaide R F, Fujimoto J G, Waheed N K, Sadda S R and Staurenghi G 2018 Optical
coherence tomography angiography [Internet] Prog. Retin. Eye Res. [cited 2018 Aug 17]
p 1–55
[36] Ido M, Osawa S and Fukukita M et al 2007 The use of colour Doppler imaging in the
diagnosis of retinal detachment Eye 21 1375–8
[37] Jia Y et al 2012 Split-spectrum amplitude-decorrelation angiography with optical coherence
tomography Opt. Express 20 4710–25
[38] Nadiarnykh O, Davidoiu V, Gräfe M G O, Bosscha M, Moll A C and de Boer J F 2019 Phase-
based OCT angiography in diagnostic imaging of pediatric retinoblastoma patients: abnormal
blood vessels in post-treatment regression patterns Biomed. Opt. Express 10 2213–26

10-19
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[39] Chu Z, Chen C L, Zhang Q, Pepple K, Durbin M, Gregori G and Wang R K 2017 Complex
signal-based optical coherence tomography angiography enables in vivo visualization of
choriocapillaris in human choroid J. Biomed. Opt. 22 1–10
[40] Arya M, Rashad R, Sorour O, Moult E M, Fujimoto J G and Waheed N K 2018 Optical
coherence tomography angiography (OCTA) flow speed mapping technology for retinal
diseases Expert Rev. Med. Devices 15 875–82
[41] Topcon O C T 2020 Angiography with the DRI Swept Source OCT Triton. Accessed Date:
11 September. Available from: http://topconmedical.com/products/ssoctangiotm.htm
[42] Mastropasqua R, Di Antonio L and Di Staso S et al 2015 Optical coherence tomography
angiography in retinal vascular diseases and choroidal neovascularization J. Ophthalmol.
2015 343515
[43] Choi J, Kwon J, Shin J W, Lee J, Lee S and Kook M S 2017 Quantitative optical coherence
tomography angiography of macular vascular structure and foveal avascular zone in
glaucoma PLoS One 12 e0184948
[44] Watzke R C, Klein M L and Hiner C J et al 2000 A comparison of stereoscopic fluorescein
angiography with indocyanine green video angiography in age-related macular degeneration
Ophthalmology 107 1601–6
[45] de Carlo T E, Bonini Filho M A, Chin A T, Adhi M, Ferrara D, Baumal C R, Witkin A J,
Reichel E, Duker J S and Waheed N K 2015 Spectral-domain optical coherence tomography
angiography of choroidal neovascularization Ophthalmology 122 1228–38
[46] Spaide R F, Fujimoto J G and Waheed N K 2015 Image artifacts in optical coherence
tomography angiography Retina 35 2163–80
[47] Rocholz R, Corvi F, Weichsel J, Schmidt S and Staurenghi G 2019 OCT angiography
(OCTA) in retinal diagnostics High Resolution Imaging in Microscopy and Ophthalmology ed
J Bille (Cham: Springer)
[48] Wei X, Hormel T T, Pi S, Guo Y, Jian Y and Jia Y 2019 High dynamic range optical
coherence tomography angiography (HDR-OCTA) Biomed. Opt. Express 10 3560–71
[49] Bressler N M and Bressler S B 2013 Neovascular (Exudative or ‘Wet’) age-related macular
degeneration Retina ed S J Ryan, S R Sadda and D R Hinton (London: Elsevier Saunders)
pp 1183–212
[50] Jia Y, Bailey S T, Wilson D J, Tan O, Klein M L and Flaxel C J et al 2014 Quantitative
optical coherence tomography angiography of choroidal neovascularization in age-related
macular degeneration Ophthalmology 121 1435–44
[51] Joshi K, KC A, Rijal R, Rai S, Lammichane G and Yadav R 2020 Compliance and visual
outcome of ‘Treat-And-Extend’ versus ‘Pro Re Nata’ dosing of intravitreal bevacizumab in
wet age-related macular degeneration—a perspective from a developing country Birat. J.
Health Sci. 5 921–26
[52] Moult E, Choi W, Waheed N K, Adhi M, Lee B and Lu C D et al 2014 Ultrahigh-speed
swept-source OCT angiography in exudative AMD Ophthalmic Surg. Lasers Imaging Retina
45 496–505
[53] Uchida A, Hu M and Babiuch A et al 2019 Optical coherence tomography angiography
characteristics of choroidal neovascularization requiring varied dosing frequencies in treat-
and-extend management: an analysis of the AVATAR study PLoS One 14 e0218889
[54] Couturier A, Mané V and Bonnin S et al 2015 Capillary plexus anomalies in diabetic
retinopathy on optical coherence tomography angiography Retina 35 2384–91

10-20
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[55] Hasegawa N, Nozaki M, Takase N, Yoshida M and Ogura Y 2016 New insights into
microaneurysms in the deep capillary plexus detected by optical coherence tomography
angiography in diabetic macular edema Invest. Ophthalmol. Vis. Sci. 57 OCT348–55
[56] Hwang T S, Jia Y, Gao S S, Bailey S T, Lauer A K and Flaxel C J et al 2015 Optical
coherence tomography angiography features of diabetic retinopathy Retina 35 2371–6
[57] Moore J, Bagley S, Ireland G, McLeod D and Boulton M E 1999 Three dimensional analysis
of microaneurysms in the human diabetic retina J. Anat 194 89–100
[58] Parravano M, De Geronimo D, Scarinci F, Virgili G, Querques L, Varano M, Bandello F
and Querques G 2018 Progression of diabetic microaneurysms according to the internal
reflectivity on structural OCT and visibility on OCT angiography Am. J. Ophthalmol. 198
P8–16
[59] Rodrigues T M, Marques J P, Soares M, Simão S, Melo P and Martins A et al 2019 Macular
OCT-angiography parameters to predict the clinical stage of nonproliferative diabetic
retinopathy: an exploratory analysis Eye [Internet] 33 1240–7
[60] Karti O, Ipek S C and Saatci A O 2020 Multimodal imaging characteristics of a large retinal
capillary macroaneurysm in an eye with severe diabetic macular edema: a case presentation
and literature review Med. Hypothesis Discov. Innov. Ophthalmol 9 33–7
[61] Khatri A, Karki P, Joshi S, Khatri B K, Kharel M, Banstola A and Shrestha S M 2020
Macular profile of eyes developing macular holes in cases of central retinal vein occlusion
treated with Bevacizumab J. Kathmandu Medical College 9 13–9
[62] Tsai G, Banaee T, Conti F F and Singh R P 2018 Optical coherence tomography
angiography in eyes with retinal vein occlusion J. Ophthalmic Vis. Res 13 315–32
[63] Khatri A, Timalsena S, Gautam S and Kharel M 2019 Varicella retinal vasculopathy:
unilateral cilioretinal artery occlusion despite acyclovir therapy caught using optical
coherence tomography-angiography (OCTA) Case Reports in Ophthalmology Medicine ed
S G Schwartz (London: Hindawi)
[64] Novais E A and Waheed N K 2016 Optical coherence tomography angiography of retinal
vein occlusion Dev. Ophthalmol 56 132–8
[65] Çelik T, Bilen F, Yalçındağ F N and Atilla H 2018 Optical coherence tomography
angiography in branch retinal artery occlusion Turk. J. Ophthalmol 48 150–4
[66] Bonini Filho M A et al 2015 Optical coherence tomography angiography in retinal artery
occlusion Retina 35 2339–46
[67] Ferrara D, Waheed N K and Duker J S 2016 Investigating the choriocapillaris and choroidal
vasculature with new optical coherence tomography technologies Prog. Retin. Eye Res 52
130–55
[68] Jia Y, Wei E, Wang X, Zhang X, Morrison J C and Parikh M et al 2014 Optical coherence
tomography angiography of optic disc perfusion in glaucoma Ophthalmology 7 1322–32
[69] Akil H, Huang A S, Francis B A, Sadda S R and Chopra V 2017 Retinal vessel density from
optical coherence tomography angiography to differentiate early glaucoma, pre-perimetric
glaucoma and normal eyes PLoS One 12 e0170476
[70] Yarmohammadi A, Zangwill L M and Manalastas P I C et al 2018 Peripapillary and
macular vessel density in patients with primary open-angle glaucoma and unilateral visual
field loss Ophthalmology 125 578–87
[71] Zhang X, Dastiridou A and Francis B A et al 2017 Comparison of glaucoma progression
detection by optical coherence tomography and visual field Am. J. Ophthalmol. 184 63–74

10-21
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[72] Wu Z, Saunders L J, Daga F B, Diniz-Filho A and Medeiros F A 2017 Frequency of testing

to detect visual field progression derived using a longitudinal cohort of glaucoma patients
Ophthalmology 124 786–92
[73] Tranos P, Karasavvidou E M, Gkorou O and Pavesio C 2019 Optical coherence tomography
angiography in uveitis J. Ophthalmic Inflamm. Infect. 9 21
[74] Kim A Y, Rodger D C and Shahidzadeh A et al 2016 Quantifying retinal microvascular
changes in uveitis using spectral-domain optical coherence tomography angiography Am. J.
Ophthalmol. 171 101–12
[75] Khairallah M, Abroug N and Khochtali S et al 2017 Optical coherence tomography
angiography in patients with behçet uveitis Retina 37 1678–91
[76] Khatri A, Wagle B, Hony K C, Chaurasiya B D, Timalsena S and Singh K et al 2021 Post
typhoid fever neuroretinitis with serous retinal detachment and choroidal involvement: a case
report Am. J. Ophthalmol. Case Rep. 21 101025
[77] Khatri A, Timalsena S and Khatri B K et al 2020 A rare entity: sympathetic ophthalmia
presumably after blunt trauma to the phthisical eye and optical coherence tomography
angiography metrics to monitor response to treatment Clin. Case Rep. 8 149–54
[78] Agrawal R, Salman M, Tan K-A, Karampelas M, Sim D A and Keane P A et al 2016
Choroidal vascularity index (CVI)—a novel optical coherence tomography parameter for
monitoring patients with panuveitis? PLoS One 11 e0146344
[79] Zahid S, Chen K C and Jung J J et al 2017 Optical coherence tomography angiography of
chorioretinal lesions due to idiopathic multifocal choroiditis Retina 37 1451–63
[80] Cheng L, Chen X and Weng S et al 2016 Spectral-domain optical coherence tomography
angiography findings in multifocal choroiditis with active lesions Am. J. Ophthalmol. 169
145–61
[81] Klufas M A, Phasukkijwatana N and Iafe N A et al 2017 Optical coherence tomography
angiography reveals choriocapillaris flow reduction in placoid chorioretinitis Ophthalmol.
Retin 1 77–91
[82] Phasukkijwatana N, Iafe N and Sarraf D 2017 Optical coherence tomography angiography
of A29 birdshot chorioretinopathy complicated by retinal neovascularization Retin. Cases
Brief Rep. 11 S68–72
[83] Prada D, Harris A, Guidoboni G, Siesky B, Huang A M and Arciero J 2016 Autoregulation
and neurovascular coupling in the optic nerve head Surv. Ophthalmol 61 164–86
[84] Garhöfer G, Chua J, Tan B, Wong D, Schmidl D and Schmetterer L 2020 Retinal
neurovascular coupling in diabetes J. Clin. Med 9 2829
[85] Warner R L, de Castro A and Sawides L et al 2020 Full-field flicker evoked changes in
parafoveal retinal blood flow Sci. Rep. 10 16051
[86] Aschinger G, Schmetterer L, Fondi K, Bata A, dos Santos V A, Garhofer G and
Werkmeister R M 2016 Quantification of total retinal blood flow via dual-beam bidirectional
Doppler optical coherence tomography for the assessment of neurovascular coupling in the
human retina Invest. Ophthalmol. Vis. Sci. 57 4619
[87] Cho K A, Rege A and Jing Y et al 2020 Portable, non-invasive video imaging of retinal
blood flow dynamics Sci. Rep. 10 20236
[88] Zhang Y S, Mucollari I, Kwan C C, Dingillo G, Amar J, Schwartz G W and Fawzi A A
2020 Reversed neurovascular coupling on optical coherence tomography angiography is the
earliest detectable abnormality before clinical diabetic retinopathy J. Clin. Med 9 3523

10-22
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[89] Nesper P L, Lee H E, Fayed A E, Schwartz G W, Yu F and Fawzi A A 2019 Hemodynamic

response of the three macular capillary plexuses in dark adaptation and flicker stimulation
using optical coherence tomography angiography Invest. Ophthalmol. Vis. Sci. 60 694–703
[90] Fondi K, Bata A M, Luft N, Witkowska K J, Werkmeister R M and Schmidl D et al 2018
Evaluation of flicker induced hyperemia in the retina and optic nerve head measured by
Laser Speckle Flowgraphy PLoS One 13 e0207525

10-23
 IOP Publishing

Photo Acoustic and Optical Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
Ayman El-Baz and Jasjit S Suri

Chapter 11
Optical coherence tomography angiography:
principles and clinical application
Ogugua Ndubuisi Okonkwo and Adekunle Olubola Hassan

Optical coherence tomography angiography (OCTA) is an imaging technology that

evolved due to the faster scan speed of available optical coherence tomograms.
Using this technology dramatically enhances understanding of the retinal vascular
network in health and several disease states. Since it is non-invasive, easily
reproducible, and provides three-dimensional images of the retinal, choroidal and
peripapillary network, several clinicians and researchers have investigated its use in
diseases such as diabetic retinopathy, macular degeneration, retinal vascular
occlusion, and glaucoma. We provide an easy-to-understand review of the basic
principles of OCTA, its clinical application, and unmet needs. An assembly of
OCTA images, with corresponding en face and cross-sectional OCT images, is
provided to illustrate the synergy between information on the retinal and choroidal
structure and vascular provided by OCT and OCTA, respectively.

Abbreviations

AMD: Age-related macular degeneration

BRM: Bruch’s membrane
CNV: Choroidal neovascularization
DCP: Deep capillary pexus
DR: Diabetic retinopathy
ERM: Epiretinal membrane
FAZ: Foveal avascular zone
FA: Fluorescein angiography
FFA: Fundus fluorescein angiography
ILM: Internal limiting membrane
INL: Inner nuclear layer

doi:10.1088/978-0-7503-2060-3ch11 11-1 ª IOP Publishing Ltd 2021

 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

IPL: Inner plexiform layer

ICP: Intermediate capillary plexus
ICGA: Indocyanine green angiography
MH: Macular hole
NFLVP: Nerve fiber layer vascular plexus
OCT: Optical coherence tomography
OCTA: Optical coherence tomography angiography
OCTARA: OCT angiography ratio analysis
OPL: Outer plexiform layer
PED: Pigment epithelial detachment
PCV: Polypoidal choroidal vasculopathy
RBC: Red blood cell
RPE: Retinal pigment epithelium
RVO: Retinal vein occlusion
RAO: Retinal artery occlusion
RNFL: Retinal nerve fiber layer
SSADA: Split spectrum amplitude decorrelation algorithm
SS: Swept-source
SD: Spectral domain
SCP: Superficial capillary plexus
VCSEL: Vertical-cavity surface-emitting lasers
VEGF: Vascular endothelial growth factor

11.1 Introduction
Advances in OCT imaging and existing age-long drawbacks of conventional dye
angiography in imaging the retina and choroid using fundus fluorescein angiography
(FFA) and indocyanine green angiography (ICGA) have resulted in the develop-
ment of OCTA, and its rapid evolution and widespread acceptance as an ophthalmic
diagnostic tool [1, 2]. OCTA offers a non-invasive, fast to produce, easily repeatable,
reproducible, and non-dye technology for assessing retinal vasculature and blood
flow within the retina’s microvasculature, optic disc, and other ocular structures in
health and several disease states [1, 3]. In summary, OCTA detects motion within the
blood vessel lumen by measuring the variation in reflected OCT signal amplitude
between consecutive cross-sectional B-scans. It does this by using laser light
reflectance of the surface of moving red blood cells to depict vessels accurately [4].
This field has undergone rapid improvements with application in several areas,
including ophthalmology and retinal medicine, research, and clinical practice [2, 4].
We present a review of OCTA’s principle and its clinical uses. In some instances we
include OCTA images, with accompanying en face and cross-sectional OCT images.

11.2 Principles of OCTA

11.2.1 Basics
OCTA evolved from Doppler based imaging of flow within the vasculature [4, 5].
With the technological advancements in newer OCT, such as the swept-source (SS)
OCT, with Fourier domain features, which produce faster A-scans due to more

11-2
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

rapid scanning speeds and enhanced axial resolution, three-dimensional OCT

images became possible [1, 2, 4].
OCTA relies on using the OCT signal variation caused by moving particles, such
as red blood cells (RBCs), as the contrast mechanism for imaging blood flow [4]. The
backscattering between moving RBCs within the blood vessels contrasts from
backscattering from a static, more stationary interface. Simultaneously, the fixed
interface signal remains stable, while the RBC signal continually changes. This
difference results in a motion contrast signal which is detectable [6].
Moving particles can be differentiated from static tissue by repeated scans
performed at the same location. Temporal changes of the OCT signal in subsequent
scans caused by the moving particles generate the angiographic contrast, enabling
visualization of the microvasculature.
In summary, In OCTA, multiple repeated B-scans are performed in the same
location, and the structural images are compared on a pixel-by-pixel basis to detect
signal changes that occur because of flowing erythrocytes. A motion contrast image
is seen as the difference between these rapidly repeated B-scans and is displayed. B-
scans consist of multiple A-scans. The rate of A-scans in available OCT scans ranges
from 70 000 to 100 000 scans per second [7]. The time interval between two B-scans
at the same scan location is known as the interscan time. The interscan time plays a
crucial role, reducing motion sensitivity and reducing motion artifacts. OCTA is a
rapidly evolving imaging technology and has gained popularity due to dye’s non-use
and eliminating all the attendant complications of dye usage [1]. Therefore, unlike
conventional dye-based angiography, OCTA can be repeated with reasonable
safety.

11.2.2 Image segmentation

This is the process of partitioning a digital image into multiple segments. The goal of
segmentation is to simplify and change an image’s representation into something
more meaningful and more comfortable to analyze. In OCTA, as in some other
digital or computer acquired images, segmentation is used to locate boundaries
(lines and curves) [8]. In general, commercial OCTA software separates the retinal
vasculature into a ‘superficial’ and ‘deep’ layer. The software predefined the retinal
layers included in these retinal slabs, used for the output images, as shown in
figure 11.1. The user has little input in this decision, and re-segmenting the retinal
layers is a laborious and often impossible task without custom analysis software [8].
The segmented OCTA provides information about the superficial and deep
vascular plexus, outer retina, Bruch’s membrane, and the choriocapillaris. The
segmentation varies between the various OCTA machines. For example, the
OptoVue AngioVue TM system designates the vasculature as either superficial or
deep plexus depending on the layer boundaries. The superficial slab or ‘plexus’ is
located between the inner limiting membrane (ILM) and the posterior border of
the inner plexiform layer (IPL), as shown in figure 11.1. The deep ‘plexus’ is the
capillaries between the posterior boundary of the IPL and the rear boundary of the
outer plexiform layer (OPL). Similarly, the choriocapillaris is the layer of capillaries

11-3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.1. Illustrates segmentation of an OCTA image into component layers, superficial and deep plexus,
outer retina, and choriocapillaris. The cross-sectional OCT image shows the layers of segmentation of the
component slabs, which are superficial plexus (ILM—IPL), deep plexus (IPL—OPL), outer retina (OPL—
BRM), and choriocapillaris (BRM—BRM + 30 microns).

in a 30 μm thick section located immediately posterior to the RPE—BRM [9]. One

can appreciate the entire segmented layers of a standard OCT in figure 11.1.
The segmentation feature seen in OCTA is an advantage over FFA, which images
only in two dimensions and demonstrates pathology in the superficial vascular
plexus [10]. In contrast, the segmentation of OCTA images allows for 3D
reconstruction and precise localization of pathology. For example, figure 11.2
represents an eye harboring an active choroidal neovascular membrane (CNVM),
and one can quickly identify this in the sub RPE location (a type 1 CNVM). The
sequential OCT shows a pigment epithelial detachment (PED) in the same area.

11.2.3 En face imaging

‘En face’ means ‘facing forward.’ In this case, the OCT acquires an image cube of the
retina using a dense raster scan. Computer software is then used to reconstruct C
scan images on the coronal plane. This computer program allows the view of a
specific retinal layer as a transverse image. Whereas traditionally, most clinicians are
used to cross-sectional OCT scans, the en face image views the retina in a coronal
plane, with the entire surface visible to the clinician all at once, figure 11.2. This type
of OCT permits a more accurate mapping of pathological and structural changes in
the examined layer. Therefore, computer software reconstruction of 3D images of
OCT images is the basis of en face imaging. The reader can identify en face OCT
images in figure 11.2. This computer technology is incorporated into both OCT
technology and OCTA, as seen in figure 11.2. Its usefulness is further illustrated
when combined with the segmentation features of the OCTA. OCTA image grading
is performed on these en face image slabs obtained by segmentation of the different
retinal layers into various horizontal depth-resolved slabs, including superficial
and deep retinal plexus, the choriocapillaris, and choroid. In this manner, it can be
used for clinical assessment and evaluating treatment response to therapeutics.

11-4
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.2. An eye with an active choroidal neovascular membrane (CNVM) in the sub RPE location (a type 1
CNVM). The en face and cross-sectional OCT image further localizes the CNVM and identifies a pigment
epithelial detachment (PED) harboring this CNVM. En face and cross-sectional images allow for precise
localization of lesions within specific subretinal layers, using their axial location on OCT cross-sections.

High-quality en face images offer a broader assessment area at the segmentation

level and enable timely diagnosis while reducing to a minimum missed pathology.
Some of the advantages of en face imaging include the ability to precisely localize
lesions within specific subretinal layers, using their axial location on OCT cross-
sections. Also, it offers the ability to register projected OCT images to other fundus
imaging modalities, using retinal vessels as landmarks.

11.3 OCTA technology: software and hardware

Several algorithms exist for using repeated B-scan OCT images to calculate motion
contrast. OCTA algorithms can use the OCT signal amplitude, the OCT signal
phase, or both amplitude and phase (known as complex amplitude) to detect motion
perpendicular to the OCT beam’s direction. The most widely used OCTA algo-
rithms use the OCT signal amplitude.
An algorithm called split spectrum amplitude decorrelation algorithm (SSADA)
was demonstrated to improve the signal-to-noise ratio and reduce sensitivity to bulk
eye motion [11, 12]. The commercially available Angiovue software on the Avanti
RTVue XR device (Optovue, Fremont, CA) uses the SSADA software. Other
algorithms in use include the optical micro-angiography (OMAG) algorithm,
[13, 14], which uses both phase and intensity information from sequential OCT
B-scans (used in the Zeiss Cirrus HD-OCT 5000 (Zeiss Meditec, Dublin, CA) with

11-5
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Angioplex) [14]. The OCT angiography ratio analysis (OCTARA) algorithm uses
intensity information to be more sensitive to low blood flow [15].
SSADA was the first algorithm to be extensively used in a commercially available
ophthalmic OCTA instrument. Decorrelation is a mathematical function that
quantifies variation without being affected by the average signal strength (as long
as the signal is strong enough to predominate over optical and electronic noise).
SSADA uses signal processing methods to divide the OCT spectrum into multiple
narrow-band spectra, which reduces the axial image resolution to minimize
sensitivity to eye motion and match the transverse OCT image resolution. Speckle
decorrelations are calculated on a B-scan-to-B-scan basis between the split spectral
data and then combined to generate a single data set with an increased signal-to-
noise ratio.
There has been rapid commercial development of OCTA in recent years. Optovue
introduced the first commercial OCTA product, the AngioVue, based on the SD-
OCT model. Topcon introduced the first commercial SS-OCTA as the Atlantis and
Triton product line, which image at 1050 nm wavelength and 100 000 A-scans per
second. Zeiss introduced the AngioVue OCTA using the SD-OCT platform and the
AngioPlex OCTA using SS-OCT. A wide array of OCTA algorithms and display
methods exist for various OCTA technologies, varying significantly between differ-
ent instrument manufacturers. Therefore, it is reasonable to exercise caution when
comparing different OCTA instruments’ results. The development of commercially
available and easy to use software and hardware meant that more clinicians could
use OCTA, which resulted in an acceleration in this technology’s growth.
The development of vertical-cavity surface-emitting lasers (VCSELs) enabled
imaging speeds of 400 000 A-scans per second, suggesting that higher imaging
speeds will be commercially available in the future [1, 16].

11.4 FA versus OCTA comparison

Though conventional FA can image only the superficial vascular plexus, it has
played a significant role in diagnosis and disease monitoring for several years and
has not been replaced by OCTA. The introduction of OCTA offers quick imaging of
retinal vasculature method that FA or ICGA methods cannot image. OCTA images
the superficial vascular plexus and shows similar patterns seen with FA in this layer
(see figures 11.3(a) and (b)). OCTA allows imaging of the intermediate, deeper, and
choroidal vascular layers. OCTA can also image the radial peripapillary vascular
network of vessels (figure 11.4). The rapid acquisition of images by OCTA means
that it can obtain serial photos of the same pathology several times, e.g., daily
imaging or several times in a day. This was not possible with FA. Therefore, OCTA
offers a wealth of clinical information from following the evolution of pathology or
monitoring the outcome of interventions.
OCTA shows superiority over FA and ICG in providing more explicit images of
vascular pathology, e.g., neovascular network, since there is no dye leakage and
therefore no hyperfluorescence to obscure the neovascularization (figure 11.5). Due
to this fact, one can visualize high-resolution, well-defined microvasculature borders.

11-6
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

(a) (b)

Figure 11.3. (a) OCTA image of left eye central retinal artery occlusion with significant macular non-
perfusion, ischemia, and perifoveal vasculature absence. There is a grossly enlarged FAZ in the deep capillary
plexus. The OCTA image shows similarity to that seen in (b), which is a late venous phase of the same eye’s
conventional fluorescein angiography.

Also, the tissue segmentation seen with OCT is possible with OCTA. Volumetric
data can undergo segmentation to visualize the vascular network at different layers
of the retina, choriocapillaris, and vascular pathology such as neovascularization.
The pathology’s actual depth or layer is seen in cross-sectional OCT images, as
shown in figure 11.2, where a subretinal pigment epithelium (sub-RPE) choroidal
neovascular membrane (CNVM type 1) is apparent.
OCTA is accompanied by the sequential acquisition of structural OCT, which
includes cross-sectional images that help co-localize pathology.
A summary of the comparison of OCTA over FFA and ICG includes the
following.
OCTA can be acquired in seconds. Since there is no need for intravenous injection
of dye, there is no nausea, vomiting or anaphylaxis, and other life-threatening
adverse events.
As OCTA is performed with considerable ease, patients have frequent follow-up
scans done.
OCTA can detect vascular abnormalities based on depth and vascular patterns
present above the RPE (type II) and between the Bruch’s membrane and the RPE
and beneath the RPE (type I).
OCTA affords image segmentation, which helps the three-dimensional assess-
ment of pathology.
A disadvantage of OCTA is that it cannot assess leakage since it is dyeless. Since
dye leakage and staining do not occur in OCTA, the boundaries and areas of
capillary drop out and neovascularization can be more precisely measured since
there are no obscurations of the borders or edges of the neovascular network
(figure 11.5).

11-7
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.4. OCTA of the radial peripapillary capillaries (RPC) and includes the vessel density computation, a
useful metric/parameter assessed in a glaucomatous eye.

Unlike conventional angiography, which is two-dimensional, making it challeng-

ing to distinguish vascular abnormalities within different layers, OCTA’s three-
dimensional nature allows for an entirely separate evaluation of retinal and
choroidal circulation.

11-8
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.5. Significant foveal and macular ischemia and areas of capillary dropout are evident on both
superficial and deep capillary plexus of this retina diagnosed with diabetic retinopathy. Also seen is a cluster of
neovascularization elsewhere (NVE) in the lower right segment of the superficial and deep slabs. Since there is
no dye leakage and therefore no hyperfluorescence to obscure, the neovascularization (NVE) outline is better
visualized than in conventional FA. The vessel density computation shows a marked reduction.

FA cannot consistently image the radial peripapillary vascular plexus, which

OCTA is very good at imaging, figure 11.4. FA underestimates the capillary density
compared to histological measurements.

11.4.1 Limitations of OCTA

OCTA cannot image leakage as can be done in conventional angiography since it is
dyeless imaging for only a short interval, in seconds. Therefore, OCTA cannot assess
changes in vascular permeability or leakage. Also, widefield FA availability, which
enables imaging of vasculature within the retina periphery, sets a high standard for
OCTA, which images a limited retina area for now. Therefore, for these reasons,
OCTA cannot replace FA. It requires faster scanners to enable the rescanning of the
same retinal position multiple times, which is the fundamental requirement under-
lying the motion contrast principle of OCTA compared to FA. A much faster scan
for the scanning of more expansive areas of retinal vasculature is required.
Therefore, the speed of the scan needed to scan a 9 mm × 9 mm or 12 mm ×
12 mm area of the retina is much faster than is necessary for a 3 mm × 3 mm retinal
scan. The initial smaller size 3 mm × 3 mm and 6 mm × 6 mm scans were limitations,
but 12 mm × 12 mm and expectations of widefield scans are more recent advances in
OCTA.

11-9
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

In fast-flow situations, the saturation limit for the OCTA will be reached;
therefore, fast flow can be detected by OCTA, but differences in flow will not be
distinguishable because they are above the saturation limit.

11.4.2 A summary of OCTA artifacts

Shadow-graphic flow projection artifact makes interpretation of en face angiograms
of deeper vascular beds more difficult. The projection artifacts occur due to the
replication of superficial vessels on deeper layers. Also, motion artifacts can occur due
to eye movements’ effect on OCTA data acquisition [17]. Patients or persons with
poorer fixation and uncontrolled eye movements are generally harder to scan. Real-
time eye-tracking and post-processing software are used to remove these motion and
projection artifacts [16]. Furthermore, to reduce noise, improve image quality, and
vessel continuity, averaging of multiple en face OCTA images is done [18, 19].
Segmentation artifacts are encountered during imaging of diseases in which there
is a significant alteration to the appearance and shape of the retinal layers. Examples
of this clinical situation are macular edema, large pigment epithelial detachment,
and bullous fluid collection in the subretinal or sub RPE space, as in figure 11.6.
Fading OCT and flow signal in large vessels due to the interferometric fringe
washout effect is associated with rapid blood flow.

11.5 OCTA imaging of the normal retina and optic disc

The retina consists of neural tissue, vascular tissue within a connective tissue
framework. OCT has revolutionized imaging, therefore understanding the retina
in health and disease states. While OCT provided information on the retina
structure, OCTA has given information on the retina’s vascular network.
Research using OCTA has provided much needed quantitative information or
data on the retina’s vasculature in a state of health and has also validated
information on vasculature obtained from a histological specimen of the retina.
The retinal vascular network in the human eye axially is divided into four distinct
capillary plexuses. While the vasculature within each plexus is densely linked,
interconnecting vessels between these sub-networks are relatively few in comparison.
With OCTA, the following layers have been confirmed and validated using confocal
microscopy ex vivo [20]. From the anterior boundary of the retina to more posterior
axial locations, the four distinct plexuses are nerve fiber layer vascular plexus
(NFLVP), superficial vascular plexus (SVP), intermediate capillary plexus (ICP),
and deep capillary plexus (DCP) [6]. While the three deeper layers can be observed
and axially separated even in the retina periphery using OCTA at sufficient axial
resolution, the NFLVP is most pronounced at locations where the nerve fiber layer
holds a substantial width, this is in the peripapillary region as well as parafoveally.
Therefore, the perifoveal vascular plexus is visualized using OCTA to assess
glaucomatous optic neuropathy. To image the healthy and diseased eye properly,
segmentation of the retinal layers is crucial in OCTA. It has been proposed that to
allow for more accurate segmentation of retinal boundaries, the retinal layer
interface between the inner plexiform layer (IPL) and inner nuclear layer (INL)

11-10
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.6. While the superficial vascular plexus is well imaged, the deep plexus cannot be easily imaged due
to difficulty with automated segmentation by the OCTA machine. This difficulty in image segmentation is
because of the large amount of subretinal fluid elevating the macula.

when shifted anteriorly and posteriorly by appropriate constant distances, represents

a conveniently defined separating boundary for visualizing the three deeper plexuses
independently within en face projections [6]. Currently, conflicting definitions of the
axial location of borders between retinal vascular networks make the direct
comparison of en face images from different OCTA devices difficult [21].
We will discuss the clinical application of OCTA next.

11-11
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

11.6 OCTA clinical application

11.6.1 Uveitis
OCTA finds usefulness in determining ischemia associated with posterior uveitis
cases. There is an absence of perfusion seen on OCTA in such uveitic eyes having
ischemia. Some uveitis cases, such as tuberculous choroiditis, are associated with
choroidal neovascularization (CNV), which can be clearly defined using OCTA. Upon
treatment with antiVEGF, sequential OCTA is used for monitoring response to
treatment. Furthermore, there is associated macular edema in several uveitis cases
evaluated using OCTA. The OCTA can find usefulness in assessing for non-perfusion
related to the CME (figure 11.7), which is associated with some cases of posterior
uveitis. Retinal angiomatosis proliferations can complicate TB granulomas, responsive
to intravitreal antiVEGF [22]. Lastly, in uveitis involving the outer retina and choroid,
patterns of flow voids in the choriocapillaris have been documented.

11.6.2 Diabetic retinopathy

OCTA has been studied extensively in evaluating DR alone and with DME. On
OCTA, one can recognize the following microaneurysms, areas of impaired
capillary perfusion (non-perfusion), defects in size and shape of the FAZ, neo-
vascularization on the disc and retina, and intraretinal microvascular anomalies
(IRMA). OCTA has been used in DR and DME clinical studies to quantify the
FAZ area, vessel density, and perfusion density in the superficial capillary plexus

Figure 11.7. This represents the OCTA and OCT of a case with chronic, severe uveitis and cystoid macular
edema. The superficial plexus does not show a distinct FAZ but suggests adequate perfusion. The deep plexus
shows a much-reduced FAZ. The absence of foveal ischemia in the presence of macular edema offers a better
visual prognosis.

11-12
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[2, 23–25]. OCTA has demonstrated a clinically significant increase in the size of
FAZ in eyes with progressively severe DR (figure 11.8(a)). OCTA is not associated
with leakage and, therefore, can distinctively visualize and define neovascularization
on the disc (NVD) and elsewhere (NVE). In this way, it is possible to differentiate
NVD from collaterals. Also, OCTA is very useful in monitoring response to
treatment [26]. It can be repeated several times in a day and at frequent clinic visits,
unlike conventional FA, and therefore detects a longitudinal change in normal and
abnormal retinal vasculature over time [26]. OCTA is also useful in identifying
macular ischemia in the presence of macular edema, which helps offer a prognosis
before commencing macular edema treatment. The finding of significant changes in
FAZ and capillary vessel density is used to monitor eyes with DR, and these findings
are more severe with the progression of DR severity. OCTA is useful in monitoring
capillary non-perfusion areas in the superficial and deep plexus, as observed outside
the fovea. Figure 11.8(b) shows the OCTA of an eye that has been treated with pan-
retinal laser photocoagulation for active PDR and received several intravitreal
antiVEGF injections for the treatment of DME. There is an enlarged FAZ seen in
the superficial plexus and significant areas of reduced vessel density and non-
perfusion in the superior perifoveal aspect of the deep plexus.
There are practical challenges experienced with OCTA imaging in the setting of
DR. In DR, when OCTA imaging is performed in the presence of significant macular
edema, image segmentation can be a challenge and result in errors using automated
segmentation by the machines (figure 11.9). Image quality, which can be a challenge in
such a clinical situation, may also affect the segmentation. Some OCTA machines
offer manual segmentation, which is preferred in this clinical situation.

11.6.3 Macular degeneration

OCTA has found usefulness in age-related macular degeneration (AMD), wet and dry
variants as well as in other forms or variants of macular degeneration. Advanced
forms of dry AMD feature flow voids, which are areas of absent flow, are visualized in
the choriocapillaris. Flow voids are better visualized using SS-OCT compared to
spectral domain (SD) OCT [27–30]. Figure 11.10 shows some areas of flow void,
which are seen as minute areas of void present within the choriocapillaris layer. The
cross-sectional OCT of this same eye shows PED and reveals drusenoid deposits seen
on en face OCT imaging. The superficial and deep plexus are within normal limits.
One of the several benefits of OCTA has been imaging neovascularization in
AMD. OCTA provides useful, clear, depth-resolved images of CNV. It is used to
study subclinical ‘quiescent’ CNV. Furthermore, OCTA has helped monitor and
understand the response of CNV to treatment with intravitreal antiVEGF [31–33].
We have used OCTA to evaluate CNV in AMD, PCV (figures 11.11(a)–(d)), and in
association with Angioid Streaks, (figure 11.12). Studies in myopic CNV have also
been done using OCTA. OCTA gives valuable information on treatment outcome,
as illustrated in figure 11.11(a–d), which shows CNV before and after treatment with
intravitreal Aflibercept. A valuable use of OCTA is its ability to localize the depth of
the CNV and therefore enable typing of the CNV. Figure 11.2 is a type 1 CNV.

11-13
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.8. (a) Shows area of capillary drop out and non-perfusion, which extends into the FAZ in the
superficial plexus; therefore, the FAZ is enlarged in this slab. A similar finding is noted in the deep capillary
plexus, but the FAZ is smaller in size. The cross-sectional OCT shows a structural thinning of the inner retina
in the area of non-perfusion. This is a case of diabetic retinopathy that presented with a featureless retina.
(b) There is foveal perfusion, with normal vasculature in the inferior foveal area. The superior foveal
vasculature and FAZ appear disorganized due to the cystic retina and hard exudates seen on en face and
cross-sectional OCT.

11-14
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.8. (Continued.)

Conversely, there is a presence of neovascularization in the outer retina and

choriocapillaris scans in figure 11.11(a–d), indicative of a type 3 CNV.

11.6.4 Retinal vein occlusion

OCTA has found usefulness in RVO for assessing areas of retinal impaired perfusion,
vascular changes such as dilatation and tortuosity, collateral vessels, cotton wool spots,

11-15
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.9. This is a clinical situation in which there is significant intraretinal and sub-retinal fluid,
distortion of the macula seen on cross-sectional OCT. This presentation can result in challenges for
automated machine segmentation and in segmentation artifacts. This segmentation difficulty can negatively
impact the quality of the OCTA. The quality of this OCTA image is reduced to a score of 7. The superficial
and deep plexus show macular non-perfusion and irregular and widened FAZ in diabetic retinopathy. Also,
there are notable hyporeflective areas in the choriocapillaris layer; these are due to projection artifacts.

Figure 11.10. There are minute hyporeflective spots seen as areas of flow void within the choriocapillaris
slab. This can be a feature of dry AMD. PED and drusen are noted on the cross-sectional and en face OCT
images.

11-16
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.11. OCTA allows for monitoring the neovascular complex response to intravitreal injection of anti-
vascular endothelial growth factors. (a) Neovascular complex at initial presentation; neovascular components
are seen in the outer retina and choriocapillaris slabs. (b) Poor response to intravitreal injection of
Bevacizumab; the neovascular complex was noted to increase in size compared to its size at initial presentation.
(c) Improved response after switching treatment to monthly intravitreal Aflibercept; the components of the
neovascular complex were noted to reduce in size significantly. (d) There is an increase in the neovascular
complex’s size following an increase of intravitreal Aflibercept treatment interval to every three months.

11-17
Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.11. (Continued.)

11-18
Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.11. (Continued.)

11-19
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.11. (Continued.)

retinal hemorrhages, increase in the areas of FAZ indicating foveal ischemia and
neovascularization [34, 35]. The increasing severity of CRVO and BRVO is associated
with decreased vessel density in both superficial and deep capillary plexus. Several
RVO studies show a decrease in both superficial and deep capillary plexus density,
FAZ enlargement, and microvascular abnormalities, as shown in (figures 11.13(a)–(c)).
RVO is mostly associated with macular edema. As indicated earlier, macular edema, if

11-20
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.12. (a) Residual choroidal neovascular complex after multiple intravitreal Bevacizumab injections.
The patient suffers from Angioid Streaks which can present with a subretinal neovascular membrane, macular
hemorrhage, and scarring, as shown in the fundus photograph (b). Fundus photograph of the left eye showing
macular scarring as a result of recurrent exudation and hemorrhage from CNVM, treated with multiple
intravitreal antiVEGF. Notice the narrow irregular streaks radiating from the disc.

11-21
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.12. (Continued.)

present, affects OCTA image quality, and segmentation in the presence of macular
edema becomes challenging. Figure 11.9 illustrates this difficulty in segmentation.
Widefield FA provides valuable information concerning peripheral ischemia in RVO,
which OCTA cannot do presently. It is hoped that SS-OCTA will offer peripheral
imaging and equal widefield FA in the future. As in the case of DR, in the setting of
enlarged FAZ and ME, it may indicate a negative prognostic outcome.

11.6.5 Retinal artery occlusion

RAO has been evaluated using OCTA. OCTA can demonstrate areas of vaso-
occlusion and non-perfusion [36–38]. In one case report, non-perfusion on OCTA
was similar to FA [39], figures 11.3(a) and (b). OCTA hasn’t been evaluated in RAO
as in RVO because of the lower incidence of RAO. Likely, the role of OCTA in
evaluating RAO will evolve further. It presents a non-invasive method of acquiring
frequent high-resolution images of the affected vasculature required in this disease,
which is investigated for more effective treatment in those non-responsive to
immediate anterior chamber paracentesis.

11.6.6 Vitreomacular interface disease

OCTA has been used for imaging pathology in the vitreoretinal interface. Epiretinal
membrane (ERM) and macular holes (MHs) have been imaged using OCTA. One can
see ERM with folds on the retina and tortuosity of the vessels shown in figure 11.14(a)
due to the tractional forces on the retinal vasculature and inner retina. A change to less
vessel tortuosity will be seen after surgery to remove the ERM. FAZ has been studied
before and after macular hole repair and suggests a reduction in FAZ area after the
successful macular hole closure [40, 41]. Figure 11.14(b) shows OCTA of MH.

11-22
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.13. (a) Shows the OCTA of a right eye that has received intravitreal anti-vascular endothelial
growth factor and scatter retinal laser (b) for treating superior-temporal branch vein occlusion. The superficial
and deep segments of the OCTA show non-perfusion in the supero-temporal quadrant and tortuosity of the
vessels in this area. Also, the vessel density in this region is reduced. Collateral vessels which can be appreciated
in the fundus photograph (b), can also be seen in the superficial and deep slabs of the OCTA. Fundus photo of
a right eye with superior branch vein occlusion. Notice the ghost vessels, collateral circulation and retinal laser
scars. (c) Shows inferior branch retinal vein occlusion, OCTA demonstrating vascular non-perfusion area with
inferior extension into the FAZ in the superficial slab.

11-23
Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.13. (Continued.)

Figure 11.13. (Continued.)

11-24
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.14. (a) Notice tortuosity and traction on the retinal vessels, which depicts the presence of ERM seen
on the deep capillary plexus of the OCTA. The en face and cross-sectional OCT scans highlight the ERMs
nicely. In this case, the FAZ in the deep plexus is irregular and appears smaller than usual because of the
ERM-induced traction, further approximating the foveal vessels closer to one another. (b) Shows the OCTA of
a full-thickness macular hole (FTMH) with normal-sized FAZ and macular perfusion.

11-25
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.14. (Continued.)

11.6.7 Glaucomatous optic neuropathy (GON)

OCTA in established glaucoma showed a reduction in perfusion in the optic nerve
head, peripapillary retina, which suggests that OCTA can be a useful addition in the
investigations for the early detection and monitoring of glaucoma. Already imaging
of the peripapillary vessel density using OCTA shown in figure 11.4, (which is not

11-26
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

(a) (b)

Figure 11.15. (a) Shows the peripapillary vessel density of a normal-appearing right eye, and one should
contrast this with the marked reduction in peripapillary vessel density of the left eye (b) of the same patient.
This peripapillary vessel density is used as an objective metric for monitoring disease progression in glaucoma
eyes.

imaged using FA) has been used to monitor glaucoma progression with high
reproducibility and sensitivity for glaucoma progression [42–44].
The disc flow index and the optic disc’s vessel density are used as two reproducible
metrics obtainable using OCTA to evaluate GON [45]. A correlation has been
observed between perfusion parameters and glaucoma stage severity, RNFL thick-
ness, ganglion cell complex thickness, and visual field mean deviation in open-angle
glaucoma patients. Other authors have confirmed this finding. Significantly reduced
peripapillary flow index and peripapillary vessel density have been demonstrated
in glaucomatous eyes compared to normal eyes, with vital diagnostic accuracy,
repeatability, and reproducibility. Figures 11.15(a) and (b) depict an OCTA of
a normal eye and a glaucomatous eye of the same patient. Notice a significant
difference in the peripapillary vessel density between the expected normal and
glaucomatous eye.
The peripapillary vessel density’s diagnostic ability, particularly the inferotem-
poral sector, in both primary open-angle glaucoma and primary angle-closure
glaucoma is of similar diagnostic accuracy to the RNFL.

11.6.8 Miscellaneous
OCTA has also been used to image other vascular and non-vascular retinal diseases,
including sickle cell retinopathy [46, 47] (figure 11.16) and retinal dystrophies such as
retinitis pigmentosa [48], and for monitoring the outcome of the surgery [49].

11.7 Discussion
Undoubtedly OCTA is innovation at its best and has provided a quantitative
assessment of retinal and choroidal vasculature in a way that has not been done
before. Since one can make image acquisition in seconds, it gives the possibility

11-27
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 11.16. OCTA of an eye diagnosed to have proliferative sickle cell retinopathy (PSCR). This
proliferative disease manifests mostly as a peripheral disease with the formation of peripheral sea fan
neovascularization. However, an enlarged FAZ can be seen on OCTA, as demonstrated in this case. This
could be indicative of perifoveal vaso-occlusion that may occur in some of these eyes.

for multiple imaging and, therefore a unique opportunity to understand vascular

pathology and interventions in more detail. Yet, there are several drawbacks to
this technology.
Currently, OCTA is not useful in the study of peripheral vascular pathology in the
way that conventional FA has been. Peripheral vascular imaging using OCTA is a

11-28
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

significant unmet need since several diseases, including retinal vein occlusions, sickle
cell retinopathy, and other peripheral vascular pathology, are currently investigated
using widefield angiography. Expected advances in OCT technology, with the
availability of faster SS scans, may soon offer widefield OCTA. Physicians appear
to prefer images acquired using SS-OCTA over SD-OCTA [22]. Preference for
images using faster scanning OCTs means that the future of OCTA may tilt more
towards the SS device.
OCTA cannot evaluate dynamic changes such as leakage, pooling, and staining,
which are essential features seen with conventional angiography, FA, and ICG. It
seems rational to expect that OCTA use will be limited in conditions in which these
dynamic phenomena will be evaluated; preferably, FA will be used instead.
Therefore, OCTA and FA can be complimentary in evaluating several retinal
pathologies.
Commercially available OCTA machines have different software that resolves
images and process the artifacts and noise generated during scanning and image
acquisition. One can expect that there will be further improvements in current
OCTA software to make the process even faster and image segmentation more
accurate and precise. Efficient removal of artifacts will enable even better imaging
and analysis of even deeper layers of the choroid and provide useful information
concerning disease effect. Yet OCTA offers an enormous amount of data that needs
resolving. One of the unmet needs in OCTA technology is the standardization of the
data across different imaging platforms. As has been suggested, there is a need to
standardize the flow within imaged vascular tissue, such as choroidal neovascular
membranes (CNVs).
Furthermore, in the clinical application of OCTA technology, there is a need for
OCTA metrics standardization. OCTA metrics or OCTA analytics refer to
numerical parameters that describe the vascular network structure as obtained
using OCTA. Some of the metrics that have been proposed for definition and
standardization include perfusion density, vessel density, blood flow index, ske-
letonized density, vessel length density, and fractal dimensions, which are all being
studied. However, several have not achieved standardized nomenclature yet.
Available OCTA technology must necessarily achieve further improvements in
OCTA image quality, accurate software segmentation, and standardization before
usage in clinical trials for retinovascular diseases. Researchers have argued that
OCTA should evolve into providing fast and reliable 3D representations and
quantification of data to provide for the vessel image interpretation.
Despite these limitations of OCTA, it is expected that this technology’s evolution
will benefit the appreciation of retinovascular pathology and herald the use of ocular
intravascular surgical techniques. The ability to provide three-dimensional image
cubes will help the accuracy of localization of retinal blood vessels. As the evolution
of OCT to intraoperative devices, i.e., intraoperative OCT (iOCT), it is likely OCTA
will someday be used during retinal endovascular surgery. Combining OCTA and
other imaging techniques such as adaptive optics (AO) will provide even greater
image resolution and understanding of intravascular blood flow [50].

11-29
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

11.8 Conclusion
OCTA is a rapidly evolving technology that has gained acceptance within the
ophthalmic community. It has benefited from advances in OCT, especially the
availability of faster scans. Hardware and software refinements will likely see even
more improvements in this technology. Standardization of software between the
various available technologies will improve widespread utility and extrapolation of
data from one source to another. Like most recent advances in ophthalmology,
patient care will improve significantly over the coming years through OCTA’s
benefits.

References
[1] de Carlo T E, Romano A and Waheed N K et al 2015 A review of optical coherence
tomography angiography (OCTA) Int. J. Retin. Vitr. 1 5
[2] Spaide R F, Fujimoto J G, Waheed N K, Sadda S R and Staurenghi G 2018 Optical
coherence tomography angiography Prog. Retin. Eye Res. 64 1–55
[3] Fang P P, Harmening W M, Müller P L, Lindner M, Krohne T U and Holz F G 2016
Technische Grundlagen der OCT-Angiographie [Technical principles of OCT angiography]
Ophthalmologe. 113 6–13 (in German)
[4] Kashani A H, Chen C L, Gahm J K, Zheng F, Richter G M, Rosenfeld P J, Shi Y and Wang
R K 2017 Optical coherence tomography angiography: A comprehensive review of current
methods and clinical applications Prog. Retin. Eye Res. 60 66–100
[5] Maram J, Srinivas S and Sadda S 2017 Evaluating ocular blood flow Indian J. Ophthalmol 65
337–46
[6] Rocholz R, Corvi F, Weichsel J, Schmidt S and Staurenghi G 2019 OCT Angiography
(OCTA) in Retinal Diagnostics ed J Bille High Resolution Imaging in Microscopy and
Ophthalmology (Cham: Springer)
[7] Borrelli E, Parravano M, Sacconi R, Costanzo E, Querques L, Vella G, Bandello F and
Querques G 2020 Guidelines on optical coherence tomography angiography imaging: 2020
focused update Ophthalmol. Ther. 9 697–707
[8] Coscas G, Lupidi M and Coscas F 2016 Image analysis of optical coherence tomography
Angiography Dev. Ophthalmol 56 30–6
[9] Agemy S A et al 2015 Retinal vascular perfusion density mapping using optical coherence
tomography angiography in normals and diabetic retinopathy patients Retina 35 2353–63
[10] Spaide R F, Klancnik J M Jr and Cooney M J 2015 Retinal vascular layers imaged by
fluorescein angiography and optical coherence tomography angiography JAMA Ophthalmol
133 45–50
[11] Jia Y et al 2012 Split-spectrum amplitude-decorrelation angiography with optical coherence
tomography Opt. Express 20 4710–25
[12] Chalam K V and Sambhav K 2016 Optical coherence tomography angiography in retinal
diseases J. Ophthalmic Vis. Res 11 84–92
[13] Zhi Z, Qin W, Wang J, Wei W and Wang R K 2015 4D optical coherence tomography-based
micro-angiography achieved by 1.6-MHz FDML swept source Opt. Lett. 40 1779–82 001779
[14] Rosenfeld P J, Durbin M K, Roisman L, Zheng F, Miller A, Robbins G, Schaal K B and
Gregori G 2016 ZEISS Angioplex™ spectral domain optical coherence tomography
angiography: technical aspects Dev. Ophthalmol 56 18–29

11-30
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[15] Stanga P E, Tsamis E, Papayannis A, Stringa F, Cole T and Jalil A 2016 Swept-source
optical coherence tomography Angio™ (Topcon Corp, Japan): technology review Dev.
Ophthalmol 56 13–7
[16] Liu L, Gao S S, Bailey S T, Huang D, Li D and Jia Y 2015 Automated choroidal
neovascularization detection algorithm for optical coherence tomography angiography
Biomed. Opt. Express 6 3564–76
[17] Spaide R F, Fujimoto J G and Waheed N K 2015 Image artifacts in optical coherence
tomography angiography Retina 35 2163–80
[18] Uji A, Balasubramanian S and Lei J et al 2018 Multiple enface image averaging for
enhanced optical coherence tomography angiography imaging Acta Ophthalmol 96 820–7
[19] Uji A, Balasubramanian S, Lei J, Baghdasaryan E, Al-Sheikh M and Sadda S 2017 Impact
of multiple en face image averaging on quantitative assessment from optical coherence
tomography angiography images Ophthalmology 124 P944–52
[20] Chan G, Balaratnasingam C, Yu P K, Morgan W H, McAllister I L, Cringle S J and Yu D Y
2012 Quantitative morphometry of perifoveal capillary networks in the human retina Invest.
Ophthalmol. Vis. Sci. 53 5502–14
[21] Rocholz R, Teussink M M, Dolz-Marco R, Holzhey C, Dechent J, Tafreshi A and Schulz S
2018 SPECTRALIS optical coherence tomography angiography (OCTA): principles and
clinical applications Heidelberg Engineering Academy https://academy.heidelbergengineer-
ing.com/course/view.php?id=505
[22] Agarwal A, Grewal D S, Jaffe G J, Stewart M W, Srivastava S and Gupta V 2018 Current
role of optical coherence tomography angiography: expert panel discussion Indian J.
Ophthalmol 66 1696–9
[23] Tey K Y, Teo K, Tan A C S, Devarajan K, Tan B, Tan J, Schmetterer L and Ang M 2019
Optical coherence tomography angiography in diabetic retinopathy: a review of current
applications Eye Vis. (Lond.) 6 37
[24] Khadamy J, Abri Aghdam K and Falavarjani K G 2018 An update on optical coherence
tomography angiography in diabetic retinopathy J. Ophthalmic Vis. Res 13 487–97
[25] Gildea D 2019 The diagnostic value of optical coherence tomography angiography in
diabetic retinopathy: a systematic review Int. Ophthalmol 39 2413–33
[26] Falavarjani K G, Iafe N A, Hubschman J-P, Tsui I, Sadda S R and Sarraf D 2017 Optical
coherence tomography angiography analysis of the foveal avascular zone and macular vessel
density after anti-VEGF therapy in eyes with diabetic macular edema and retinal vein
OcclusionOCTA After Intravitreal Injection Invest. Ophthalmol. Vis. Sci. 58 30–4
[27] Arya M, Sabrosa A S, Duker J S and Waheed N K 2018 Choriocapillaris changes in dry age-
related macular degeneration and geographic atrophy: a review Eye Vis. (Lond.) 5 22
[28] Moreira-Neto C A, Moult E M, Fujimoto J G, Waheed N K and Ferrara D 2018
Choriocapillaris loss in advanced age-related macular degeneration J. Ophthalmol 2018
8125267
[29] Waheed N K, Moult E M, Fujimoto J G and Rosenfeld P J 2016 Optical coherence
tomography angiography of dry age-related macular degeneration Dev. Ophthalmol 56
91–100
[30] Müller P L, Pfau M, Schmitz-Valckenberg S, Fleckenstein M and Holz F G 2020 OCT-
angiography in geographic atrophy Ophthalmologica 244 42–50

11-31
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[31] Perrott-Reynolds R, Cann R, Cronbach N, Neo Y N, Ho V, McNally O, Madi H A,

Cochran C and Chakravarthy U 2019 The diagnostic accuracy of OCT angiography in naive
and treated neovascular age-related macular degeneration: a review Eye (Lond) 33 274–82
[32] Malamos P, Tsolkas G, Kanakis M, Mylonas G, Karatzenis D, Oikonomopoulos N,
Lakoumentas J and Georgalas I 2017 OCT-angiography for monitoring and managing
neovascular age-related macular degeneration Curr. Eye Res 42 1689–97
[33] Coscas G, Lupidi M, Coscas F, Français C, Cagini C and Souied E H 2015 Optical
coherence tomography angiography during follow-up: qualitative and quantitative analysis
of mixed type I and II choroidal neovascularization after vascular endothelial growth factor
trap therapy Ophthalmic Res 54 57–63
[34] Tsai G, Banaee T, Conti F F and Singh R P 2018 Optical coherence tomography
angiography in eyes with retinal vein occlusion J. Ophthalmic Vis. Res 13 315–32
[35] Moussa M, Leila M, Bessa A S, Lolah M, Abou Shousha M, El Hennawi H M and Hafez T
A 2019 Grading of macular perfusion in retinal vein occlusion using en-face swept-source
optical coherence tomography angiography: a retrospective observational case series BMC
Ophthalmol 19 127
[36] Baumal C R 2016 Optical coherence tomography angiography of retinal artery occlusion
Dev. Ophthalmol 56 122–31
[37] Bonini Filho M A et al 2015 Optical coherence tomography angiography in retinal artery
occlusion Retina 35 2339–46
[38] de Castro-Abeger A H, de Carlo T E, Duker J S and Baumal C R 2015 Optical coherence
tomography angiography compared to fluorescein angiography in branch retinal artery
occlusion Ophthalmic Surg. Lasers Imaging Retin. 46 1052–4
[39] Okonkwo O N, Hassan A O, Akanbi T, Umeh V C and Ogunbekun O O 2021 Vitrectomy
and manipulation of intraocular and arterial pressures for the treatment of nonarteritic
central retinal artery occlusion Taiwan J. Ophthalmol. [Epub ahead of print]
[40] Wilczyński T, Heinke A, Niedzielska-Krycia A, Jorg D and Michalska-Małecka K 2019
Optical coherence tomography angiography features in patients with idiopathic full-thick-
ness macular hole, before and after surgical treatment Clin. Interv. Aging 14 505–14
[41] Kita Y, Inoue M, Kita R, Sano M, Orihara T, Itoh Y, Hirota K, Koto T and Hirakata A
2017 Changes in the size of the foveal avascular zone after vitrectomy with internal limiting
membrane peeling for a macular hole Jpn. J. Ophthalmol 61 465–71
[42] Chansangpetch S and Lin S C 2018 Optical coherence tomography angiography in glaucoma
care Curr. Eye Res 43 1067–82
[43] Kwon H J, Kwon J and Sung K R 2019 Additive role of optical coherence tomography
angiography vessel density measurements in glaucoma diagnoses Korean J. Ophthalmol 33
315–25
[44] Chung J K, Hwang Y H, Wi J M, Kim M and Jung J J 2017 Glaucoma diagnostic ability of
the optical coherence tomography angiography vessel density parameters Curr. Eye Res 42
1458–67
[45] Yarmohammadi A et al 2016 Optical coherence tomography angiography vessel density in
healthy, glaucoma suspect, and glaucoma eyes Invest. Ophthalmol. Vis. Sci. 57 OCT451–9
[46] Alam M, Thapa D, Lim J I, Cao D and Yao X 2017 Quantitative characteristics of sickle cell
retinopathy in optical coherence tomography angiography Biomed. Opt. Express 8 1741–53

11-32
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[47] Han I C, Linz M O, Liu T Y A, Zhang A Y, Tian J and Scott A W 2018 Correlation of ultra-
widefield fluorescein angiography and OCT angiography in sickle cell retinopathy
Ophthalmol. Retin. 2 599–605
[48] Mastropasqua R et al 2020 Widefield swept source OCTA in retinitis pigmentosa Diagnostics
(Basel) 10 50
[49] Okonkwo O N, Sibanda D, Akanbi T and Hassan A O 2020 Natural history of a large
bubble of migratory submacular perfluorocarbon Int. Med. Case Rep. J. 13 477–86
[50] Camino A, Zang P, Athwal A, Ni S, Jia Y, Huang D and Jian Y 2020 Sensorless adaptive-
optics optical coherence tomographic angiography Biomed. Opt. Express 11 3952–67

11-33
 IOP Publishing

Photo Acoustic and Optical Coherence Tomography Imaging,

Volume 3
Angiography: an application in vessel imaging
Ayman El-Baz and Jasjit S Suri

Chapter 12
A comprehensive survey on computer-aided
diagnostic systems in diabetic retinopathy
screening
Meysam Tavakoli and Patrick Kelley

Diabetes mellitus (DM) can lead to significant microvasculature disruptions that

eventually cause diabetic retinopathy (DR), or complications in the eye due to
diabetes. If left unchecked, this disease can increase over time and eventually cause
complete vision loss. The general method to detect such optical developments is
through examining the vessels, optic nerve head, microaneurysms, haemorrhage,
exudates, etc from retinal images. Ultimately this is limited by the number of
experienced ophthalmologists and the vastly growing number of DM cases. To
enable earlier and efficient DR diagnosis, the field of ophthalmology requires robust
computer aided diagnosis (CAD) systems. Our review is intended for anyone, from
student to established researcher, who wants to understand what can be accom-
plished with CAD systems and their algorithms to modeling and where the field of
retinal image processing in computer vision and pattern recognition is headed. For
someone just getting started, we place a special emphasis on the logic, strengths and
shortcomings of different databases and algorithms frameworks with a focus on very
recent approaches.

Important acronyms
• Diabetes mellitus: DM
• Diabetic retinopathy: DR
• Non-proliferation diabetic retinopathy: NPDR
• Proliferation diabetic retinopathy: PDR
• Microaneurysm: MA
• Haemorrhage: HE

doi:10.1088/978-0-7503-2060-3ch12 12-1 ª IOP Publishing Ltd 2021

 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

• Exudate: EXs
• Cotton wool spots: CWS
• Neovascularization: NV
• Macular edema: ME
• Field-of-view: FOV
• Computer-aided diagnosis: CAD
• Optic nerve head: ONH
• Receiver operating characteristic: ROC
• Area under the curve: AUC
• k-nearest neighbor: kNN
• Support vector machine: SVM
• Neural network: NN
• Artificial intelligence: AI
• Machine learning: ML
• Deep learning: DL
• Convolution neural network: CNN
• Recurrent neural network: RNN
• Autoencoder: AE

12.1 Introduction
The rapidly rising diabetes mellitus (DM) is one of the major issues of our recent
health care [1]. The number of people affected by DM continues to increase at an
alarming rate [2, 3]. The World Health Organization estimates for the next 25 years
the number of diabetic people to grow from 130 to 350 million [4]. In the same
direction, in the time between 2000 and 2030 another source anticipates the spread of
the DM population worldwide to increase from 2.8% to 4.4% [5]. The real catch is that
just half of diabetic people are diagnosed with the disease [6]. From a medical aspect
viewpoint, DM leads to serious late complications such as macro and micro vascular
changes which cause diabetic retinopathy (DR), heart disease, and renal problems
[7, 8]. In DM, because of insulin insufficiency there is incapacitated metabolism of
glucose; this in turn leads to damage and complications in blood vessels [9]. According
to a study in the United States, DM is among the five lethal diseases, and yet there is
no treatment for it. In the same study, the total yearly expense to diagnose and treat
DM in 2002 was estimated to be $132 billion [10]. In fact, unmanaged DM and its
complication leads to different malfunctions and death.
DR is a serious issue stemming from DM that affects the eye and is a main cause
of vision loss and blindness [11, 12]. DR is the resulting condition specifically for
these complications and damaged vessels lying in the tissue behind the retina [13].
Unfortunately, so many cases of diabetes go undiagnosed and the connection
between diabetes and eye vision isn’t common knowledge; patients with DM are 25
times more likely to be at risk of vision loss when compared with non-diabetic ones
[14]. DR is often overlooked and detected too late, since it is a silent disease, and it
may be diagnosed by the patient when the vascular abnormalities in the retina have
progressed to a stage when its cure is complicated and sometimes unfeasible [15, 16].

12-2
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Therefore, the most successful therapy for DR can be managed just in its early
phases. Consequently, early detection of DR through systematic screening is of
critical importance.
It is believed that the screening of diabetic people for the evaluation of DR
potentially reduces the risk of blindness in these people by 50% [17]. The problem is,
however, that this would produce immense costs, due to the large number of
examinations. Moreover, there are not enough specialists, especially in rural areas,
to do these examinations [18]. Physical screening of DR to address the large and
mounting number of DR cases is therefore not possible. Instead, automated systems
based on image processing methods may help to overcome this issue [19]. It would be
more beneficial and helpful if the initial steps of analyzing the retinal images and
detecting all retinal lesions in the images can be computerized through a robust
computer automated system. In this way, the automated system would allow for
speedier identification of the telltale signs of DR from the early onset of retinal
lesions and notification of the ophthalmologist for further assessment. This would
allow more patients to be screened per year and for ophthalmologists to better
manage and direct their time towards those priority patients who would have
otherwise been overlooked or delayed in treating [20].

12.1.1 Clinical signs of diabetic retinopathy

Now, before we start DR diagnosis the big question is what are the revealing
indications of DR? There are several abnormalities in the retina that are manifes-
tations of DR which are concisely described below.
1. Microaneurysms (MAs): They are the first visible sign of DR. They appear as
small and round shape dots near tiny retinal blood vessels in retinal images [21].
The size of MAs usually ranges from 10 μm to 125 μm in diameter [19, 22].
2. Haemorrhages (HEs): A more severe level of DR is caused by retinal
haemorrhages or leakage of weak capillaries and ruptured MAs into the
surface of the retina [23]. They are defined as a red spot with different shapes,
such as ‘dot’, ‘blot’ and ‘flame’ [21, 23] and irregular margin and uneven
density [24].
3. Hard exudates (EXs): When the lipoproteins and other proteins are leaking
because of the degradation or breakdown of the blood–retina barrier through
abnormal retinal vessels, hard EXs appear [25]. They become visible as
small yellowish-white patches with sharp borders [24]. They are often
collocated in bulks or semicircular hoops and positioned in the exterior
layer of the retina [24].
4. Cotton wool spots (CWS): Also called soft exudates, they occur due to
occlusion, or blockage of blood supply, of arterioles and cause disruption and
damage to areas of tissue [26]. The ischemia of the retinal layer of the nerve fiber
caused by reduced blood flow to the retina changes the axoplasmic flow and
leads to accumulation of axoplasmic residue in the retinal ganglion cell axons.
The accumulation of this axoplasmic residue arises as fleecy yellowish-white
lesions in the retinal nerve fiber layer known as soft exudates or CWS [26, 27].

12-3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

5. Neovascularizations (NVs): The new blood vessels start to grow on the inner
surface of the retina primarily because of the lack of oxygen. These new
blood vessels push into the surrounding retinal space and are feeble and very
often bleed into the retinal cavity, effecting the eyesight [10, 28].
6. Macular edema (ME): It is recognized as the swelling beneath the macula, an
area at the center of the retina. It results in penetration of abnormal
capillaries in the retinal because of the fluid leakage or solutes around the
macula [29, 30]. Eventually, it influences the central the eyesight [31].

Figures 12.1 and 12.2 show typical normal retinal images with their basic
landmarks, and abnormal images with different signs and stages of DR.

12.1.2 Stages of diabetic retinopathy

According to the existence of the aforementioned clinical signs, and the severity/
density of these signs [34], DR is categorized into four kinds including: mild,
moderate, severe non-proliferative DR (NPDR), and PDR [24, 35, 36]. With respect
to number of MAs and HEs, the NPDR is introduced as normal, mild, moderate,
and severe DR [37].
• Normal: If no DR sign is observed.
• Mild: Here, only MAs are presented.
• Moderate: In this case, both MAs, and HEs exist and their numbers are less
than 20 in each quadrant, and we see soft EXs as well.
• Severe: Here, in each quadrant there are more than 20 MAs, and HEs. Also,
in more than two quadrants there are some venous beadings and different
shapes of hard EXs.

Figure 12.1. A normal retinal image versus a DR one. Sample retinal images for both normal and abnormal
ones. On the left image, all available anatomical landmarks have been shown including fovea, macula, blood
vessels, optic nerve head [32]. On the right one: all pathologic signs of DR including MAs, HEs, and EXs are
visible [33].

12-4
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 12.2. Different types of DR. (a) Normal; (b) mild NPDR; (c) moderate NPDR; and (d) severe NPDR.

Figure 12.2 shows the retinal images of these categories and related signs. In the
PDR, we see the initiation of new vessels, which are abnormal vessels (NVs). It is the
severe phase of DR, in which this new abnormal vessel growth, caused by poor
circulation especially from formations of MAs, protrudes into the retina and even
into the vitreous, the gel-like medium that gives the eye its spherical shape. Thus,
both MAs and NVs are two important clinical lesions [38] and fluid in DR is
classified as EXs and non-EXs [39].
There are several organizations for DR grading such as American Academy of
Ophthalmology, whose classification of DR was established by curing of DR at its
early stage [40], and a protocol introduced by the Scottish DR grading system [41].
Table 12.1 presents the Scottish protocol.
MAs, in general, are a very good indication of possibly worsening conditions of
DR. From a statistics point of view, the NPDR type with MAs has a 6.2% chance to
expand into PDR during a year [1]. An increased number of MAs is a primary
characteristic of DR progression. Moreover, with development of ischemia, there is
an increased chance of PDR progress during the first year. The first year risk growth
increases from 11.3% (lower) to 54.8% (advanced) phase [1, 42]. In the NVs stage,

12-5
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 12.1. DR Scottish grading protocol [41].

Grade Characteristics

No DR Without abnormalities
Mild NPDR Just MAs appear
Moderate NPDR There are more than 5 MAs but less than severe NPDR
Severe NPDR • In each quadrant, there are more than 20 HEs

• In two quadrants, there is venous beading

• Intra-retinal microvessel abnormalities

PDR • Any NVs at ONH or other parts of the retina

• Vitreous/ pre-retinal HE

No ME There are no EXs or thickening of retina in posterior pole

Mild ME There are EXs or thickening of retina at posterior pole, >1 disc diameters
from fovea
Moderate ME Same as mild ME’s signs plus 1 disc diameters, without affecting fovea
Severe ME There are EXs or thickening of retina with affecting fovea

diabetics have a 25.6% to 36.9% chance of vision loss and blindness, if not cured
appropriately. Furthermore, PDR eyes not therapied after more than two years have
a chance of vision loss and blindness at 7.0% and if for more than four years it is not
cured, the chance of vision loss goes up to 20.9%. On the other hand, this vision loss
and blindness decreases to 3.2% during two years and 7.4% during four years of
therapy [42]. Diabetic people with mild DR do not need specific treatment. In fact,
they need to manage their DM and the related risk factors such as hypertension,
anaemia, and renal malfunction. They need to be closely monitored, to decrease the
possibility of progressing to higher stages of DR [18]. In the severe and advanced
phase of DR, like NV, therapy is limited [43].

12.1.3 Retinal imaging and databases

Now we are aware of the signs and stages of DR, and the way diagnosis and
determination of the stage of DR involves processing retinal images. Retinal
photography is a procedure where three-dimensional retinal landmarks are pro-
jected on the two-dimensional image. The created image shows the amount of
reflected light [44]. In the camera that creates the fundus image, the optical layout is
equivalent to an ophthalmoscope, which is used to observe the inner part of the eye.
The camera sees the retinal area with 2.5 times magnification at the angle between
30° and 50°. More information can be found in [1]. In retinal photography for DR
screening, the below modalities/techniques are utilized [44].

12-6
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

(a) Red-free retinal imaging: The image is taken by applying the amount of
reflected light at a particular waveband.
(b) Color retinal image: This image is based on the red–green–blue (RGB)
spectrum and the camera detector sensitivity to the light.
(c) Fluorescein angiography retinal image: The created image is based on the
amount of photons that are emitted by fluorescein dye. The dye is injected
through the blood stream [44].

Automated algorithms, in the process of testing and improving upon themselves,

require a set of exemplary data with the proper classifications. In our case, these
would be retinal images from reputable databases, many of which have been
classified by professional ophthalmologists. Most of the provided retinal images in
these databases are used in application of screening and testing the CAD methods
related to detection of MAs, HEs, exudates and other DR features. Outlined below,
many of these online research databases are included. Table 12.2 summarizes all
databases for detection of DR.
• DIARTDB: The dataset was created by Kauppi et al [45] and is freely
available for research from http://www.it.lut.fi/project/imageret/. Images were
taken with a 50° field-of-view (FOV). It has two classes, DIARTDB0,
including 130 color retinal images (20 images are normal without sign of
DR and 110 contain symptoms of the DR). These signs include hard and soft
EXs, MAs, HEs and NV. The second class is called DIARTDB1, including
89 images, of which five are normal and 84 contain at least mild NPDR signs
(MAs).
• HRF: It is presented as the High-Resolution Fundus retinal image dataset.
The images were captured with different image acquisition settings and 45°
FOV. There are images in this database for each category: healthy, glau-
comatous, and DR. Information in the database was created by a group of
experts from ophthalmology clinics [46].
• STARE: It consists of 400 images initially applied by Hoover et al [47]. These
images were captured at 35° FOV.
• KAGGLE: The images were prepared by Kaggle [37] as a series of high
resolution pictures. The images were captured under different spatial con-
ditions and were indexed by a well-experienced ophthalmologist with respect
to without DR to mild, moderate, severe and proliferative DR.
• ROC: Retinopathy Online Challenge is a dataset established for enabling
research groups to establish a CAD system for diagnosis of DR. It aimed to
create a platform for algorithms evaluation [48].
• DRIVE: It has been developed from a DR screening system. The database
included 40 retinal images in which 33 images are without symptoms of DR
and seven showing signs of DR. The database was taken using a non-
mydriatic 3CCD fundus camera with a 45° FOV [49].
• MESSIDOR: It consists of 1200 eye retinal color images. The images were
captured by a color non-mydriatic retinography with a 45° FOV with
different resolutions [50].

12-7
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 12.2. More details for the databases used for DR detection system.

Database Numbers Resolution Format Task

MESSIDOR [50] 1200 1440 × 960, TIFF DR grading

2240 × 1488, Risk of DME
2304 × 1536

Kaggle [54] 80 000 — JPEG No sign of DR-Mild

Moderate-Severe-PDR

E-ophtha [55] 148 MAs, 2544 × 1696 JPEG MA small HE detection

233 normal, 1440 × 960 EX detection
47 EXs,
35 no signs of DR

DRIVE [49] 33 normal 584 × 565 TIFF Vessels extraction

7 mild DR stage

STARE [47] 402 605 × 700 PPM 13 retinal diseases

Vessels segmentation-ONH

DIARETDB1 [45] 5 no sign 1500 × 1152 PNG MAs-HEs

84 NPDR
symptom

CHASE [37, 56] 28 1280 × 960 JPEG Vessels extraction

MUMS-DB [19, 120 PNG MAs

53]
ARIA [57] 16 normal 768 × 576 TIFF ONH-Fovea-Vessel
92 AMD-59 DR

SEED-DB [58] 192 normal 3504 × 2336 — ONH

43 glaucomatous

EyePACS [59] 9963 — — Referable DR

MA

ORIGA [60] 482 normal 720 × 576 — ONH

168 glaucomatous

• E-OPTHIA: It was created because of a DR screening project which

developed a color dataset for DR, called E-ophtha. This dataset was made
of fundus images with a variety of signs such as EXs and MAs manually
labelled by well-experienced ophthalmologists [51]. It contained 47 images
with EXs, 35 with no sign, and 281 images with MAs or small HEs.

12-8
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 12.3. Some CAD systems and databases (Blue = B, Green = G, Red = R).

Image
Author plan Method Database

Pourreza et al [53] B&G Radon transformation DRIVE, STARE, MUMS-DB

Esmaeili [61] R&G Morphology STARE, DRIVE, DIARETDB1
Qureshi et al [62] G Entropy and Hough DIARETDB0, DIARETDB1,
DRIVE
Welfer et al [63] LUV Mathematical Morphology DRIVE, DIARETDB1
Harangi and Hajdu Grey Probabilistic model and DIARETDB0, DIARETDB1,
[64] augmented Naive based DRIVE, MESSIDOR
Akyol et al [65] G BLP DIARETDB1, DRIVE,ROC
Dai et al [66] Grey Hough, PCA MESSIDOR, DRIONS
Alshayeji et al [68] Grey Adaptive edge detection, DERIVE, DMED, STARE,
DIARETDB1
Fraz et al [68] G Watershed transformation DRIVE, Shifa, CHASEDB1,
DIARETDB1

• DRIONS: An evaluating dataset for ONH segmentation that included 110

retinal images. The fundus images were captured with a color analogical
camera [37].
• CHASEDB1: It is a dataset available at blogs.kingston.ac.uk/retinal/cha-
sedb1 for free. It includes 28 fundus images. An ophthalmologist and human
experts were asked to mark the vessels in all these images and these hand-
labelled vessels were used as the ground truth (or gold standard) for this
dataset [52].
• HEI-MED: The database is a test database for evaluating algorithms in the
EXs and macular edema detection. It contained 169 images which were
segmented manually by a highly skilled ophthalmologist [30].
• Others: These are other datasets established by individual researcher groups
such as MUMS-DB [19, 53] DMED, ARIA, Eyepack [1]

Table 12.3 summarizes some CAD systems using different image database
analysis for DR detection. These approaches are distinguished based on features,
processing, and classification of DR.

12.2 Computer-aided diagnosis and diabetic retinopathy

As previously mentioned, manual detection and diagnosis is an exhaustive task in
both cost and lack of experts [16] when screening DR. It would be more cost
effective and helpful if the initial task of analyzing the retinal images could be
automated. Automated approaches address these issues by decreasing the time,
expense, and effort considerably [69]. Moreover, since image processing techniques

12-9
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

are growing in all areas of medical science, by assisting them, especially in advanced
ophthalmology, we can do automated screening [39].
For this reason, automated DR detection and classification utilizes computer
aided diagnostic (CAD) systems. These computer-based systems have the ability to
detect any change in normal and abnormal images and systematize these variations
to form a feature space. At the end, the mixture of these features introduces type and
stage of DR. Different CAD systems have been presented as the state-of-the-art for
early DR detection and related lesions [12, 18, 19, 70–77].
Normally, each CAD system is the sequence of two different approaches, i.e.,
feature detection and classification [37]. Since there are many different approaches in
CAD systems, it is also necessary to evaluate their robustness and accuracy. There
are many different standard methods used in assessing the effectiveness of CAD
systems. A common one is using receiver operating characteristic (ROC) analysis
and concept of area under the curve (AUC) analysis. The AUC of ROC curve is a
measurement of performance for an automated method, for example, extraction
problem, at different thresholds values. ROC is a probability curve and AUC
shows degree or rate of distinguishability. In fact, AUC shows us how much the
method is able to distinguish between classes. The higher the AUC, the better is
the method in its prediction. By analogy, the higher the AUC, the better the
method is at distinguishing between images with DR and no DR. ROC curves
illustrate the tradeoff between sensitivity and specificity for a range of thresholds
and enable the identification of an optimal value [78].
Here, the analytical definition when using ROC is assessed according to the true
positive fraction (TPF), given by sensitivity, and the false positive fraction (FPF) (1-
specificity) [16]. Moreover, the accuracy is specified as a quantification including the
ratio of well-classified pixels. The final outputs for the automated approach as
compared to the ground truth or gold standard are calculated for each image. These
metrics are defined as:
TP
Sensitivity(Se ) =
TP + FN
TN
Specificity(Sp ) = (12.1)
TN + FP
TP + TN
Accuracy(Acc ) =
TP + FN + TN + FP
where TP is true positive, TN is true negative, FP is false positive and FN is false
negative same as in [79, 80].
A look into the results for both ROC and AUC (equation (12.1)) obtained by
other CAD systems in DR screening. Table 12.4 summarizes some of the algorithms
at different image analysis stages for detection of DR.
Briefly, both Tavakoli et al [19] and Philip et al [36] introduced a DR screening
system based on the comparison of automated system with respect to the gold
standard which was manual detection by an ophthalmologist. Their approach

12-10
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 12.4. Some of the CAD systems with different image analysis stages.

CAD system Methodology Authors

ONH detection Qureshi et al [62], Welfer et al [63], Harangi and

and Hajdu [64], Akyol et al [65], Dai et al [66], Alshayeji
segmentation [67], Basit and Fraz [68], Bharkad [81], Lesay et al
[82], Zou et al [83], Xiong and Li [84], Sarathi et al
[85], Lu [86], Lowel et al [87], Hoover and
Goldbaum [88]
Vessel Thresholding Hoover et al [47], Dash and Bhoi [89], Zhu et al [90],
segmentation approach Neto et a l[91], Xu and Luo [92], Nguyen et al [93],
Tracking approach Zhang et al [94], Fraz et al [95], Marin et al [96],
Machine classifiers Akram et al [97], Vermeer et al [98], Chakraborti
et al [99], Wu et al [100], Tavakoli et al [101],
Roychowdhury et al [102], Tolias and Panas [103],
Azzopardi et al [104], Zana et al [4, 105], Soares
et al [106]
Lesion detection Red lesions (MAs, Quellec et al [107], Walter et al [22], Tavakoli et al
HEs) [19], Wang et al [108], Srivastava et al [109], Seoud
et al [23], Pereira et al [112], Dai et al [110], van
Grinsven et al [111]
Hard and soft Liu et al [113], Zhang et al [115], Orlando et al [114],
exudates Youssef and Solouma [116], Fraz et al [117], Amin
et al [118], Sopharak et al [119], Sanchez et al [120]
DR detection and DR, Non DR Tavakoli et al [19], Koh et al [121], Gardner et al[122],
classification NPDR, PDR Rubini and Kunthavai [123], Gulshan et al [124],
Kose et al [125], Yang et al [126], Sisodia et al [127],
Kumar et al [128], Gupta et al[129], Leontidis [82],
Sinthanayothin et al [130], Usher et al [131],
Fleming et al [17, 132], Perumalsamy et al [133],
Winder et al [134], Mookiah et al [135]

detected DR images with an accuracy of more than 90%. Sinthanayothin et al [130]

established a trustable automatic DR online screening system purposed to simplify
ophthalmologists’ work in their country. They got sensitivity and specificity for their
screening program to classify the DR as 80.2% and 70.7%, respectively. Usher et al
[131] employed a DR screening tool which detected DR with both sensitivity and
specificity of 95.1% and 46.3%, respectively. Niemeijer et al [70] introduced a DR
detection algorithm based on information fusion approach with AUC of 0.88.
Quellec et al [71] established an automated DR screening system for both MAs and
age-related macular degeneration detection, and their method reached an AUC of
0.93. Fleming et al [132] introduced a CAD system to detect blot HEs that provided
sensitivity and specificity of 98.60% of 95.50%, respectively. Fleming et al [17] also
established another CAD system to detect MAs which got a sensitivity and

12-11
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

specificity of 85.4% and 83.1%. Perumalsamy et al [133] employed a CAD system

and obtained an accuracy of 81.3% by evaluating the performance with ophthal-
mologists’ idea. Abramoff et al [136] established a web-based DR detection system
with a specific protocol and short questionnaire, measurement of visual eyesight,
and four retinal images. The accuracy of their protocol reached 93%. Abramoff et al
[137] also presented an automated DR detection system which acquired an AUC of
0.84 of DR detection.
There are also some authors who reviewed most of the CAD methods, and their
applications [1, 39, 44, 134, 138] for automated screening of DR. Winder et al [134]
discussed different segmentation approaches of optic nerve head (ONH), retinal
vessels, macula, and DR detection. Teng et al [138] for retinal landmarks, and
lesions segmentation, and image registration reviewed different methods. Abramoff
et al [44] surveyed various retinal imaging modalities, vessel segmentation, lesion,
and ONH detection. Patton et al [39] discussed image preprocessing, and registra-
tion methods, besides landmarks and lesions segmentation. Mookiah et al [1]
comprehensively reviewed the approaches to detect and extract the retinal lesions
and landmarks such as ONH, vessels, EXs, MAs, and HEs.
From a classification viewpoint, color characteristics were utilized on a
Bayesian classifier to group each pixel into either lesion or non-lesion categories
[139]. In this study, authors obtained 100% accuracy in detection of all EXs in the
retinal images, and accuracy of 70% in classification of normal retinal images.
By using image processing techniques and multilayer neural network (NN), DR
and normal retina were categorized with sensitivity and a specificity of 80.2%
and 70.7%, respectively [21]. Automated detection of NPDR, based on MAs, HEs,
and EXs was studied in [135, 140]. The approach was able to correctly detect the
NPDR stage with an accuracy of 81.7%. For DR screening again above, lesions
were used [141] and they obtained the sensitivity and specificity of 74.8% and
82.7%, respectively. An early automated DR detection system was developed by
Kahai et al based on a decision support system [142]. In their method, Bayes
optimality criteria were used to identify DR with a sensitivity and specificity of
100% and 67%. Different stages of DR were grouped using both area and
geometry of the combination of the vessels with a feedforward NN [143]. The
average classification efficiency for this method was 84% with both sensitivity and
specificity of 90% and 100%. Nayak et al employed EXs and vessel area along with
parameters related to its texture which was coupled with NN to classify NPDR,
PDR, and normal [5]. They got accuracy of 93%, sensitivity 99%, and specificity of
100%. The feature of support vector machine (SVM) classifier was used by
Acharya et al [144] to classify the retinal image in to normal, mild, moderate,
severe and PDR categories. They showed a classification accuracy of 82% with
sensitivity of 82%, and specificity of 88%. Nicolai et al established an automated
lesion detection algorithm for DR screening purpose, which detected 90.1% of
diabetics and 81.3% of those without DR [145]. Their system showed both
sensitivity and a specificity of 93.1% and 71.6%.

12-12
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

12.3 Retinal landmarks detection

A standard CAD screening system needs to extract the retinal landmarks such as
fovea, ONH, and vessels, and additionally lesions such as MAs, HEs, EXs, CWS,
and NVs. Several studies have worked on various detection and segmentation
approaches to identify these anatomical regions. Here, we briefly discuss them.

12.3.1 Localization and segmentation of optic nerve head

The ONH can reveal a critical eye disease like glaucoma [146]. The shape of the
ONH appears as semicircle or ellipse in the retinal image. The ONH is also
applied as a reference landmark to identify other retinal regions and structures
such as the macula and fovea [87, 147–149]. Moreover, the ONH localization
assists in determining central retinal arteries and veins [87, 150]. Many vessel
tracking algorithms use the ONH center as the starting point, since the vessels
diverge from the ONH center. The ONH localization is carried out by detecting
its center or by drawing a circle around the ONH region [151–155]. ONH
segmentation is introduced by specifying the boundary of ONH. Thus, identi-
fication and extraction of ONH is highly important [134, 146, 156–160].
The localization of the ONH is serious to differentiate it from the other lesions.
The ONH identification approaches, as explained in literature mainly estimate its
center or surround it by a circular mask. In both cases, the localization is quite an
elaborate task because of the presence of thick vessels, inaccurate boundaries, and
lesions such as EXs [134, 146]. The methods according to intensity changes are easy
to use, fast and robust for images with no symptoms. These methods may lose out
when the ONH is concealed by blood vessels and lesions such as EXs, CWS, or other
similar intensity artifacts [161].

12.3.1.1 Detection of optic nerve head center

The ONH color and shape can be utilized to detect it. Normally, in some studies the
ONH is localized by its relations to bright pixel clusters. Sinthanayothin et al [162]
used a window size 80 × 80 pixel as the size of a typical ONH and localized the ONH
by observing the maximum change in its intensity to its neighboring pixels. They
chose the point with highest pixel variance to detect the center of the ONH and
reported a sensitivity and specificity of 99.1%. Lowel et al [87] developed a
correlation filter to identify the ONH in the presence of EXs and other artifacts
and correctly detected it in 83% of cases. Hoover and Goldbaum [88] presented an
approach using the concept of a fuzzy approach to locate the ONH. Their approach
obtained 89% correct ONH detection. Walter et al [163] introduced a method to
detect ONH based on its illumination, distance, and watershed transformation in
different color space, HSL (hue, saturation, luminance), with accuracy of more than
85%. Tavakoli et al applied three different vessel segmentation algorithms,
Laplacian-of-Gaussian, Canny, and Matched filter edge detectors to detect the
ONH in both the images without DR signs and in the presence of DR [164].
The Hough and similarly the radon transform were applied by many authors to
identify the ONH [53, 87, 165–167]. The method is based on integral transformation,

12-13
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

or mapping the values of the line integral of an images’ pixels into another domain.
Ultimately, the ONH was identified by effective projection profiles in radon space
and the boundary parts with maximum vote [53].
Principal component analysis (PCA) was another method for identifying the
ONH [162, 168, 169]. In fact, this method turns the correlated variable into
uncorrelated ones, which are known as principle parts [1]. The first principal
component demonstrates the maximum information existing in the image. Li and
Chutatape [169] detected the ONH using the concept of PCA with accuracy of 99%.
Another method that applied the convergence of vessels to localize the ONH was
proposed by Foracchia et al [170]. Youssif et al [171] applied matching the expected
directional structure of the vessels for ONH detection. Their method started by
normalizing luminosity and contrast throughout the image using illumination
equalization and adaptive histogram equalization methods, respectively. Lu and
Lim [172] utilized a different approach to place the ONH based on its bright
appearance in a color retinal image. They used a series of concentric lines with
various directions and assessed the image variation along the different directions.
Figure 12.3 shows the retinal images of these categories and related signs.

12.3.1.2 Segmentation of optic nerve head

In the case of segmentation methods category, the brightness was applied by Li and
Chutatape [173, 174] to detect the ONH candidates for their model-based technique.

Figure 12.3. ONH detection. (a) and (c) retinal images from MUMS-DB and detected center of the ONH;
(b) and (d) ONH mask segmentation.

12-14
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

They established an active shape model, which consisted of building a model with
training cases and iteratively matching the landmark points on the disc edges and
main vessels inside the disc [169]. A model-based approach was introduced
effectively based on active contour model by Osareh et al [175, 176] to approx-
imately detect the ONH. Lowell et al carefully created an ONH template, which
was then correlated to the intensity component of the retinal image using the full
Pearson-R correlation [87]. Another model-based algorithm was introduced by Xu
et al [177]. They applied the deformable contour model technique that included
ONH margin. Their method used a clustering-based classification of contour points
and customized contour evolution step integrated in original snake formulation. The
approach introduced by Wong et al [178] was based on the level-set method followed
by ellipse matching to smooth the boundary of the ONH. They proposed: (1) the
ONH location according to the previously obtained location by means of histogram
analysis and initial contour definition, and (2) subsequently, a modified version of
the conventional levelset method used for ONH boundary extraction from the red
channel. This contour was finally matched by an ellipse.
On the other hand, Lu [86] introduced a circular transformation to take ONH
and its boundary simultaneously. Lalonde et al presented Hausdorff-based template
matching together with pyramidal decomposition and confidence assignment [179].
Pyramidal decomposition acts mainly for detection of large areas with bright pixels,
with probable existence of the ONH, but it can be disturbed by large areas of bright
pixels that may be near the image borders due to non-uniform illumination. In order
to solve this problem, Frank ter Haar [180] used illumination equalization in the
green channel of the image, and then a resolution pyramid using a simple Haar-
based discrete wavelet transform was formed. Further, he applied the Hough
transform and proposed two methods using a vascular branch constructed from a
binary vessel image. The last technique of Frank ter Haar was based on fitting the
vessel orientations on a directional model [180]. The ONH identification methods
not only apply the characteristics of the ONH, but also extract the location and
orientation of vessels [181–184]. Niemeijer et al [184] presented the use of local vessel
geometry and image intensity features to find the correct positions in the image. A
k-nearest neighbour (kNN) regressor was used to accomplish the integration. Tobin
et al [181] used an algorithm that highly depended on vessel-related ONH character-
istics. They applied a Bayesian classifier to categorize each pixel in images as with/
without ONH. Abramoff and Niemeijer [185] applied same properties for the ONH
detection, and then a kNN regression was used to estimate the ONH location. The
approach developed by Abramoff et al [186] according to a pixel classification using
the feature analysis and nearest neighbor technique. The final results of their
algorithm came with the categorization of pixels to a group that belonged to the
ONH and non-ONH. In a recent study, Hsiao et al placed ONH by illumination
correction operation, and contour segmentation that was completed by a supervised
gradient vector flow snake [187]. To identify ONH location candidates, first
template matching was used by Yu et al. Next, they applied vessel features on
the ONH to place the location of it. They combined region and local gradient

12-15
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

information to the segmentation of the ONH boundary and used morphological

filtering to remove retinal vessels [188].
Moreover, some approaches made use of the fact that all major vessels diverge from the
ONH [170, 171, 180, 189]. Most of the studies identified the ONH with respect to its shape
and color. Other algorithms recognized the ONH according to tracking vessels toward
their origin. In table 12.5, we have provided the details of more methods in ONH detection.

12.3.2 Vessel segmentation

There is a substantial effort reported in the state of the art for the segmentation of
retinal vessels in fundus images [56, 91, 94, 95, 97, 190–195]. Overall, the vessel
segmentation approaches are classified into five categories: tracking, mathematical
morphology, matching filter, model-based thresholding, and supervised classification
algorithms [1]. Table 12.6 summarizes all these methods and we briefly discuss them
below.

Table 12.5. Summary of some methods and their results of ONH detection.

Authors ONH detection method DRIVE STARE

Sinthabayothin et al Highest average variation 99.1% 42%

[162]
Walter and Klein [163] Largest brightest connected object 100% 58%
Li and Chutatape [173, Brightness guided, model base 99% —
174]
Osareh et al [175, 176] Averaged ONH model based 100% 58%
Lowell et al [87] ONH Laplacian of Gaussian template 99% —
Lalonde et al [156] Hausdorff based with pyramidal decomposition 100% 71.6%
and confidence assignment
Frank ter Haar [180] Resolution pyramid using Haar based wavelet 89% 70.4%
transform
Frank ter Haar [180] Hough used only to pixels on or close to the 96.3% 71.6%
retinal vasculature
Frank ter Haar [180] Pyramidal decomposition of both the vascular 93.2% 54.3%
and green plan
Frank ter Haar [180] Hough and fuzzy convergence 97.4% 65.4%
Tobin et al [181] Vasculature related ONH features and a 81% —
Bayesian classifier
Abramoff and Niemeijer Vasculature related ONH properties and a 99.9% —
[184, 185] kNN regression
Youssif et al [171] Matching the expected directional pattern of 100% 98.8%
the retinal blood vessels
Hsiao et al [187] Illumination correction, and contour 100% 90%
segmentation is completed by a supervised
gradient vector flow snake
Tavakoli et al [164, 167] Different vessel segmentation methods 90% 85%
Pourreza et al [53] Radon Transform 100% 94%

12-16
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 12.6. Some vessel segmentation methods.

Classifier and
Authors Method databases Evaluation

Xu and Luo [92] Sobel filter for ONH, Thick and thin vessels, Acc. = 93.36%
adaptive threshold SVM, DRIVE Sen. = 85.57%
Hassan et al [196] Hidden Markov Vessels, DRIVE Acc. = 95.7% Sen. = 81.0%
model Sp. = 97.0%
Dash and Bhoi [89] Binary difference and DERIVE, Acc. = 95.5%,
enhanced image CHASEDB1
Nguyen et al [93] Linear combination SVM, STARE, Acc. = 93.24%, 94.07%,
of line responses DRIVE, REVIEW
Fraz et al [56, 95] Mathematical DRIVE, STARE Acc. = 95.52% Sp. = 97.23%
morphology, MESSIDOR
aggregation
Marin et al [96] Grey level features ANN, DRIVE, Acc. = 95.26%, Sen. = 69.
STARE 44%sp. = 98.19%
Aslani and Sarnel Hybrid features DRIVE, STARE Acc. = 95.13%, Sen. = 77.
[197] 96% sp. = 97.17%
Lupascu and AdaBoost AdaBoost classifier, Acc. = 95.69%
Trucco [198] DRIVE Sen. = 55.74%
sp. = 99.36%
Rodrigues et al Histogram analysis, DRIVE, HRF Acc. = 94.65%,
[159] wavelet Hessian
based and
Gaussian filter
Tavakoli et al [101] Radon transform MUMS-DB Acc. = 90%, Sen. = 85%
Zhang et al [94] kNN DRIVE Acc. = 95.05%
Sen. = 78.12%
Sp. = 96.62%
Zhu et al [90] Vessel profile, ELM, DRIVE, RIS Acc. = 96.07, Sen. = 60.18,
Gaussian, Sp. = 98.68%
morphology
Neto et al [91] Binary DRIVE, STARE Acc. = 86.16%, 87.87%
morphological Sen. = 79.42, 76.96%
reconstruction Sp. = 95.37%, 96.31%
Vega et al [199] Lattice NN and ANN Acc. = 94.83%,
dendritic Sen. = 96.71%,
processing Sp. = 70.19%
Waheed et al [200] Connected Localized Fisher Acc. = 95.81%, 96.16%,
component, binary discriminant
threshold, analysis, DRIVE,
STARE
Zilly et al [201] AdaBoost MESSIDOR, —
DRISHTIGS

12-17
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

A comprehensive review on existing methods in retinal vessel segmentation and

available public datasets is presented in [33, 52]. In general, vessel segmentation
algorithms can be classified into two broad categories: unsupervised and supervised
methods [202]. A comparative survey of these two classes has been presented in
[102]. Unsupervised methods can be further classified into techniques based on
matched filtering [47, 99], morphological processing [203], vessel tracking, multi-
scale analysis [204, 205], line detectors [93] and model-based algorithms [98, 206].
Supervised segmentation methods are based on pixel classification such as kNN [49],
Gaussian mixture models [106], SVM [207], NN [207], decision trees [194], and
AdaBoost [96, 198]. They utilize ground truth data for the classification of vessels,
based on given features.
The principle of matched filter detection in unsupervised methods was proposed
in [47]. With this technique, the authors used a 2D Gaussian-shaped template in all
directions to search for vessel components. The kernel was rotated through a range
of angles (usually 8 or 12) to detect vessels with different orientations. The resulting
image is thresholded to produce a binary representation of the retinal vasculature.
Another matched filtering approach was used for retinal vessel segmentation [208].
The application of local thresholding strategies, edge detection, matched-filtering
and region growth was developed in [208]. Chakraborti et al [99] applied an
unsupervised segmentation approach that combined a ‘vesselness’ filter and matched
filter using an orientation histogram. Vermeer et al [98] achieved vessel detection by
thresholding, after convolving the image with a 2D Laplacian kernel. A general
framework of adaptive local thresholding based on the combination of applying a
multithreshold scheme and classification procedure to prove all binary result objects
were applied by Jiang et al [209]. Zana et al [4] proposed a morphology-based
approach using an algorithm that combined morphological filters and cross-
curvature evaluation to segment vessel-like patterns in retinal images. Other
mathematical morphology approaches are described in [105, 203, 210]. Heneghan
et al [210] combined morphological transformations with curvature information and
matched filtering for centerline detection. In [100] vessels are extracted using
multiple structure elements followed by connected component analysis. Martinez-
Perez et al [204] introduced an automated algorithm for retinal images based on a
multi-scale feature detection. The local maxima of the gradient magnitude and the
maximum principal curvature of the Hessian tensor were used in a multi-pass region
growing process. This procedure accurately segmented the retinal vasculatures by
using both feature and spatial information. Zhang et al [205] employed innovative
rotating multi-scale second-order Gaussian derivative filters which are referred to
as orientation scores for the enhancement and segmentation of blood vessels.
Roychowdhury et al proposed an iterative unsupervised retinal vessel segmentation
algorithm that used an adaptive threshold and a region growing method with a
stopping criterion to terminate the iteration [211].
Nguyen et al proposed a vessel detection method based on line detection [93]. This
approach was based on the fact that changing the length of a line detector makes line
detectors with a variety of scales. The final vessel segmentation results were achieved
by combining line responses at varying scales. The combination of shifted filter

12-18
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

responses for detection of bar-shaped structures in retinal images was presented by

Azzopardi et al [104]. Their approach was rotation invariant, where the selection of
the orientations was chosen from given structures similar to vessels which suffered
from difficult crossing cases. Lam and Yan [206] used the Laplacian operator to
segment blood vessels and detect centerlines from the normalized gradient vector
field, pruning noisy objects. Vessel continuity employs measures of width and
orientation, iteratively calculated in a local region near the current point, in order to
track along the length of a vessel [100, 103]. Tavakoli et al [101] presented the
combination of radon transformation and multi-overlapping windows to segment
the retinal vessels. Since their approach relied on the concept of radon, which is
integral based, they suppressed the intrinsic noise of images by using this trans-
formation. Figure 12.4 shows a sample example for the results of vessel
segmentation.
Supervised approaches use a pixel classification method, referred to as a
primitive-based approach by Staal et al [212]. This approach was based on the
detection of ridges, used as primitives for introducing linear segments, called line
elements. Sinthanayothin et al [162] identified retinal vessels using an NN whose
inputs were deduced from PCA of the image and edge extraction of the first principal
component. An NN and backpropagation multilayer method was proposed by
Gardner et al [122] for vessel tree detection following histogram equalization. The
kNN classifier was implemented by Niemeijer et al [49]. In their algorithm, from the
green plane of the RGB image, a vector of 31-component pixel features was created
with the Gaussian and its derivatives up to order 2 at five different scales. Soares et al
[106] applied a Gaussian mixture model Bayesian classifier. By using the Gabor
wavelet transform, multiscale analysis was performed on the image. Ricci and

Figure 12.4. Vessel segmentation. (a) and (c) Retinal images from MUMS-DB; (b) and (d) vessel segmentation
results. For more details see [101].

12-19
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Perfetti [207] utilized an SVM for classifying pixels as vessel or non-vessel. In their
algorithm, along with the gray-level of the target pixel they applied two orthogonal
line detectors to create the feature vector. One of the interesting supervised methods
was implemented by Fraz et al [56] which was an ensemble classification system of
boosted and bagged decision trees. The drawback of supervised classifications is its
necessity for a sufficient number of manually annotated training image sets.
Moreover, it is not easy to generalize the trained models to meet the needs of
different datasets. More recent studies have successfully applied the concept of deep
learning to the segmentation of the retinal vasculature [90, 213–222].

12.3.3 Localization of the fovea

The fovea is a small region approximately at the center of the retina and responsible
for central and high resolution vision. It is placed around 45 times of ONH radius to
the temporal direction of ONH [151, 162, 169, 184, 223–228]. The visual cells placed
in the fovea are closely packed and cause the highest sharpness of vision [229]. There
are not any blood vessels in the fovea region to block the crossing of light to the
fovea cells. The fovea detection approaches are divided into two classes: hybrid and
constraint based approaches [1].

12.3.3.1 Hybrid-based method

Sinthanayothin et al [162] presented a method for fovea localization using intensity
correlation. At the beginning, the fovea was identified with intensity correlation and
from HSI transformation the peak was selected. The selected peak located a dark
area that was considered as the fovea. The disadvantage of their method was that
when the fovea was not at the center it failed. In another method, using location
of the fovea, a polar retina coordinate system was established. Here, the fovea
accurately was located using its appearance and geometry according to other
landmarks [169]. Niemeijer et al [184] proposed a method to identify the fovea
based on a single point distribution model. The method included all signs including
global and regional, to identify this model. These signs were extracted from the
related vessel position and its width. Tavakoli et al [147] proposed an approach
based on vessel segmentation and parabola fitting to identify the fovea. Their

Figure 12.5. Fovea detection. (a) Retinal images from MUMS-DB; (b) vessel segmentation results; and
(c) detected fovea. For more details see [147].

12-20
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

algorithm included four steps: multi-overlapping windows, local radon transform,

vessel validation, and parabolic fitting. Figure 12.5 shows an example for the results
of fovea detection.

12.3.3.2 Positional constraints-based

The positional restrictions [162, 169, 183, 184, 223–225] were applied to localize the
fovea and determine its position. The fovea identification was obtained using the
location of other retinal regions. The main part of this approach [181] specifies
the horizontal retinal ridge that is passing across the ONH and fovea, which divides
the upper and lower retinal regions. Further, the region of fovea was extracted using
a fixed distance of 5 ONH radius from the ONH center. Table 12.7 summarizes
some related methods with brief details.

12.4 Detection of retinal lesions

An acceptable CAD system requires identifying lesions such as MAs, HEs, EXs,
CWS, and NVs as perfect as possible to finally result in distinguishing between
DR and no DR images. Several studies have worked on various detection and
segmentation approaches to detect these lesions. Here, we briefly discuss them.

12.4.1 Microaneurysm and haemorrhage detection

The most reliable method for MA detection is from fluorescein angiography retinal
images [19]. However, the fluorescein angiography images are costly, invasive and
not suitable for everyone [22]. Moreover, this approach is time-consuming.
Introductory global image processing approaches were applied for automated
detection. There are a number of methods for the detection of both MAs and
HEs. These methods can be generally categorized into two general groups including
unsupervised and supervised approaches [17, 22, 107, 108, 110, 111, 226, 233–247].
Many studies have searched both MAs and HEs detection from the retinal images
by using proper thresholding after masking the landmarks such as ONH, fovea, and
vessel. See figure 12.6, from [2], for more details. After the thresholding,

Table 12.7. Details of some fovea detection methods.

No. of
Author Approach images Accuracy

Sinthabayothin et al [162] NN and template-based 112 80.4%

Li and Chutatape [174] Modified active shape model 35 100%
Tavakoli et al [147] Fitting the main vessels with parabola function 140 90%
Tobin et al [181] ONH and fovea localization 345 92.5%
Seckar et al [231] Hough transform and morphological function 44 100%
Fleming et al [183] Vessel segmentation with semiellipse fitting 1056 96.6%
Niemeijer et al [184] Point distribution model and kNN 500 96.8%
Welfer et al [232] ROI selection and morphological operator 126 96.1%

12-21
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

mathematical morphology was used to distinguish MAs and other structures, such
as bifurcations and small vessels. From an unsupervised viewpoint, the MAs
detection approaches are classified in to different classes: morphology, region
growing, wavelet, and hybrid approaches [1]. Among these groups, the HEs detection
is mostly based on mathematical morphology combined with pixel classification. In this
section, we briefly explain the related studies.
Walter et al [22] presented a four-part method for MAs detection. At the beginning,
the preprocessing step used green channel images. Top-hat transformation and global
thresholding were used to identify MAs. Next, fifteen characteristics were used to
differentiate true MAs from false ones using kNN, Gaussian and kernel density
classifiers. Hatanaka et al [248] proposed a brightness correction approach for automated
detection of HEs. The correction was applied in hue saturation value (HSV) space.
In the morphological approach, different adjustments often increase the accuracy of
MAs detection [17, 240]. Despite this type of hand-crafted method being typically fast,
easy, and user- friendly, its ability is constrained by its builder. In better words, some main
hidden structures and uncovered patterns might be ignored by the builder and cause false
detection [110]. The morphological operations such as closing [22], or using top-hat
transformation [19] have been applied for MAs detection due to their relatively uniform
circular shape and limited size range. However, besides the above issue, most of the
mathematical morphology methods mainly depend on the choosing of optimal structur-
ing elements; changing their size and shape decreases the efficiency of these algorithms.
Fleming et al [17] developed contrast normalization, watershed and region growing
approaches to detect MAs in retinal images. Moreover, they applied a 2D Gaussian-
after median filtering to preprocess the image. To detect MAs they used the kNN
classification using nine characteristics from extracted candidates. Sinthanayothin
et al [21] identified MAs by applying region growing and adaptive intensity thresh-
olding, combined with a moat operator. Quellec et al [107] used wavelet transform
algorithm for MAs detection by matching wavelet sub-bands with the lesion template.
Tavakoli et al [19] employed an approach based on top-hat, mathematical
morphology and radon transform, applied locally in multi-overlapping windows
to detect MAs. Niemeijer et al [233] introduced a hybrid red lesion detection method
using ideas from Spencer et al [249] and Frame et al [250]. At the beginning, using
pixel classification the red lesion candidates were identified [249]. The MAs and HEs
were detected after masking the vessels. Usher et al [131] and Gardner et al [122]
applied NN to locate MAs. Initially, they used local adaptive contrast enhancement
to preprocess the image. After that, landmarks (ONH and vessel segmentation), and
red lesions (MAs and HEs) were identified using region growth and adaptive
intensity thresholding combined with moat operation [21]. Zhang et al [251]
described a method using multi-scale correlation filtering and threshold values
within angiographic retinal image for MA detection. They used moving Gaussian
kernels to extract the correlation coefficient for pixels. Larsen et al [145] introduced
an automatic MAs and HEs detection approach using classification algorithm called
RetinaLyze using shape and size of the lesions.
Fleming et al [132] applied multiscale and morphological operators to place HEs.
For the preprocessing, they corrected the variations in intensity and contrast by

12-22
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

applying median filter and histogram equalization. In another study [17], they
proposed a region growth algorithm to identify large lesion regions. In their
approach, they had difficulty differentiating HEs and MAs. Zhang and Chutatape
[252] introduced an approach to detect HEs using SVM. In addition, they applied
PCA to detect the features. These features were used as the inputs of SVM to extract
HEs. Hipwell et al [253] proposed an automated MA detection in red-free retinal
images. To remove variation in the illumination, a shading correction was applied to
the image.
Another approach in contrast with linear landmarks like vessels, which are
directional, is to utilize template matching filters with multiscale Gaussian kernels
[107, 235, 251]. According to statistical results, the intensity distribution of MAs is
matched to a Gaussian distribution [107] because MAs display Gaussian profiles in
all projections. Therefore, a template matching filter highly increased the accuracy
of MA detection. Using this idea, Quellec et al proposed a wavelet transform
method for MA detection. In their algorithm, they detected the MAs using a local
template matching filter in the wavelet domain [107]. Zhang et al [251]used a multi-
scale correlation coefficient based approach. In their method, they used a nonlinear
filter with five Gaussian kernels at various standard deviations to detect MA
candidates. Ram et al [239] described a feature based method that rejects specific
classes of clutter while accepting a higher number of true MAs. The potential MAs,
obtained after the final step, are labelled a grade that was based on their shape
similarity to true MAs. Similar to the morphological approach, this type of
algorithm is still limited by ignorance of hidden and unnoticeable structures.
Moreover, a lot of parameters are required to be assigned empirically.

Figure 12.6. MA detection. (a, d, and f) Retinal images from different datasets; (b, e, and g) vessel
segmentation and masking results; and (c, f, and h) detected MAs.

12-23
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Table 12.8. Details of some of the MA and HE detection methods.

Authors Methods Performance

MA detection
Tavakoli et al (2013) [19] Radon transform Sen. = 92%, Sp. = 90%
Hipwell et al (2000) [253] Matched filter, region Sen. = 81%, Sp. = 93%
growing
Sinthanayothin et al (2002) [21] Moat operator Sen. = 77.50%, Sp. = 88.70%
Larsen et al (2003) [145] Size and shape Sen. = 71.4%
Usher et al (2004) [131] Moat operator Sen. = 95.10%, Sp. = 46.30%
Niemeijer (2005) [233] Pixel classification using Sen. = 100%, Sp. = 87%
kNN
Fleming et al (2006) [17] Contrast normalization, Sen. = 85.40%, Sp. = 83.10%
watershed region growing
Walter et al (2007) [22] Gaussian filtering, top-hat Sen. = 88.5%
Quellec et al (2008) [107] Optimal wavelet transform Sen. = 89.62%, Sp. = 89.50%
Zhang et al (2010) [251] Multi-scale correlation Sen. = 71.30%
filtering and dynamic
thresholding
Sanchez (2011) [73] Sen. = 91%
Antal and Hajdu (2012) [254] Ensemble-based AUC= 0.90
Lazar and Hajdu (2013) [235] Cross section features —
Pereira (2014) [112] multi-agent system Sen. = 87%
approach
Seoud (2016) [23] Dynamic shape features Sen. = 84.4%
Srivastava (2017) [109] Filters for feature extraction —

HE detection
Gardner (1996) [122] NN and statistical threshold Sen. = 73.80%
tuning
Zhang and Chutatape (2005) PCA and SVM Sen. = 89.10%
[252]
Fleming et al (2008) [132] Multi-scale, morphology Sen. = 98.60%, Sp. = 95.50%
and SVM
Hatanaka et al (2008) [248] Brightness correction and Sen. = 80%, Sp. = 88%
thresholding

The brief explanation of the MAs and HEs detection algorithms is introduced
in table 12.8.

12.4.2 Exudates segmentation

Here, we briefly review the studies of EXs segmentation. EXs are one of the main signs
of DR, because they cause retinal edema and consequently vision loss and blindness.
The approaches for identification of EXs [113, 255–257] can be categorized into four

12-24
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

classes: pixel clustering, morphology, thresholding/region growing, and classification-

based approaches [1]. Here, we concisely introduce these methods.
Several studies [119, 258, 259] have used Fuzzy C-Means cluster-based method for
EXs detection. This method suggests pixels with various classes and changing degrees of
membership for segmentation purpose [260, 261]. Osareh et al [258] first by correcting
color and contrast worked on image preprocessing. The preprocessed images were used
for a segmentation step using Fuzzy C-Means. The genetic algorithm was applied to
choose characteristics from Fuzzy C-Means clustered regions [259]. The optimal
characteristics were then used in the NN to detect EXs and non-EXs regions. This
approach obtains the differing retinal color. In another study, the EXs were segmented
from low quality images using Fuzzy C-Means [119]. The same authors presented EXs
detection using Bayesian classifier [119]. Hsu et al [69] described an algorithm of EXs
segmentation using cluster-based approach. Then, the cluster-based method was
utilized to classify the abnormal pixels. The restriction of the above methods is to
know considerable clusters which contribute to the EXs detection.
Morphological [120, 262, 263] and thresholding/region growth methods [21,
135] were applied in many studies about EXs detection. Sopharak et al [262]
presented an EXs segmentation approach for low image quality by applying
adjusted morphology. Walter et al [263] detected EXs using gray level changes
of green channel. After first step detection, their boundaries were determined
applying a morphological reconstruction approach. Their method did not differ-
entiate EXs from CWS. Sanchez et al [120] used the same approach as Walter et al
[263] used in preprocessed the green channel and a histogram of the image was
modelled using a Gaussian mixture model. Furthermore, a thresholding method
was utilized to detect the EXs using Gaussian mixture model. Sinthanayothin et al
[21] developed a region growing-based method by selecting some threshold values
in the gray scale images. To improve the quality of image, adaptive local contrast
enhancement was applied. This algorithm located other retinal regions and
landmarks such as the ONH, retinal vessels and fovea. Fleming et al [183]
presented a multi-scale morphology algorithm to localize EXs. They used
Gaussian and median filter to preprocess the green channel images for correcting
the contrast variations. A watershed region growing algorithm was used to
separate the EXs from other landmarks. Finally, They used SVM to classify
regions into EXs and non-EXs. Welfer et al [264] presented an EXs detection
algorithm based on morphology operator. They applied a set of morphological
functions to detect EXs. Mookiah et al [135] introduced EXs detection by
applying color, shape and morphological processing. Their method assumed
EXs to have higher contrast than the ONH.
Several studies [122, 233, 265] utilized machine learning approaches to categorize
EXs and non-EXs pixels. Gardner et al [122] utilized NN to classify EXs applying
gray scale of retinal images. Sanchez et al [265] segmented EXs according to Fisher’s
linear discriminant analysis. They used color normalization and contrast enhance-
ment. Niemeijer et al [233] introduced an automated method for EXs detection using
classification based on a machine learning idea. The pixels with high probability
were classified into EX class from non-EXs.

12-25
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

12.5 Artificial intelligence in diabetic retinopathy

In general, understanding and controlling DR or other retinal diseases has become
extensively more complicated because of the large number of images and diagnoses.
In fact, processing all these patient examinations seems as a big data challenge [266].
Clearly, the new era of diagnosis and clinical data mining immediately needs
intelligent systems to control them sufficiently, safely and efficiently.
Recently, through the advancement of supervised learning ideas, the recent
application of artificial intelligence (AI) approaches for DR detection and its related
lesions have significantly increased in the research community. AI has already
revealed proof-of-concept in medical science such as radiology and pathology, which
like ophthalmology relies heavily on diagnostic imaging. As its turns out, image
processing is the most important application of AI in healthcare [267]. Moreover, AI
is particularly suitable in assisting clinical tasks by using efficient algorithms to
identify and learn features from large volumes of imaging data, and helping to
reduce diagnostic errors. In addition, it can recognize specific patterns or lesions and
correlate novel characteristics to obtain innovative scientific insight [268].
The most important branch of AI is machine learning (ML) [268] where the rules
are learned directly by algorithms from a series of examples, called training data,
instead of being encoded by hand. ML techniques need a set of biomarkers or
characteristics to be measured directly from the training data (e.g. labeled lesions in
the retinal image). After that, based on a training database of characteristics with
known labels, i.e., hand crafted features, a classifier learns to identify the correct
label from the newly seen characteristics. Once a few strong classifiers have been
established, the effectiveness of such ML models mostly relies on the differentiative
power of the chosen characteristics which underpin the classifier performance. There
are different reviews that have summarized ML approaches [1, 15, 74, 269–272].
Both Mookiah et al [1] and Mansour [74] classified DR diagnosis methods in respect
of affiliated methods, such as DR lesion tracking, mathematical morphology,
cluster-based, matched filter, and hybrid methods. Faust et al [15] surveyed studies
that detect lesion characteristics from retinal images, such as the retinal vessels,
MAs, HEs, and EXs. Joshi and Karule [269] summarized the first researches on EXs
segmentation. Almotiri et al [270] came up with a survey of methods to detect the
vessels. Almazroa et al [271] and Thakur and Juneja [272] studied several
approaches for ONH detection.
Recently, the accessibility of large databases and the immense computing power
suggested by advanced graphics processing units have research on deep learning
(DL) based methods, which have demonstrated magnificent efficiency in all medical
sciences. Many DL-based approaches have been established for a variety of works to
analyze images to establish automated CAD systems for detection of DR. The
relation between AL, ML, and DL has been shown in figure 12.7.

12.5.1 Overview of deep learning

There are multiple DL-based algorithms that have been presented. Some frequently
utilized deep architectures for the DR detection purpose include convolutional

12-26
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Figure 12.7. Relation between AI, ML and DL.

neural networks (CNNs), autoencoders (AEs), recurrent neural networks (RNNs)

and deep belief networks (DBNs). A complete overview of these architectures is
found in [273]. There are several DL-based architectures related to DR detection
[274–284]. Based on the clinical significance of automated DR detection, we group
applications of DL into three broad classes: (i) vessels segmentation, (ii) ONH
localization and segmentation, and (iii) retinal lesions detection and classification. In
the below sub-sections, we briefly introduce the literature about DL-based algo-
rithms for these tasks. A brief summary of the DL methods in DR detection systems
is presented in table 12.9.

12.5.1.1 Retinal blood vessel segmentation

Many of the retinal vessel segmentation approaches are based on CNN architectures
[216, 217, 285–289].
Maji et al [285] proposed 12 CNN models for this purpose. Each CNN model
included three convolutional layers and two fully connected layers. Liskowski and
Krawiec [217] introduced a supervised pixel-wise algorithm using deep CNN, which
was trained by preprocessed retinal images, and augmented using geometric trans-
formations. Tan et al [216] employed a seven-layer CNN model to segment blood
vessels, ONH, and fovea.
Some methods use AEs in segmentation of retinal vessels [290, 291]. Roy and
Sheet [290] presented an AE-based deep NN model that used the domain adaptation
method for training the model. The AE-based deep NN included two hidden layers,
which were trained using the source domain databases, using an autoencoding and
supervised mechanism. The algorithm introduced by Maji et al [285] employed a
hybrid DL, which consisted of unsupervised stacked denoising AEs with a total of
500 layers, to segment vessels in retinal images.

12.5.1.2 Optic nerve head detection

As we mentioned at the beginning of this chapter, ONH identification can improve
DR detection and classification. In fact, the ONH brightness can cause uncertainty

12-27
 Table 12.9. Details of some of the MA and HE detection methods.

Authors Methods Training Type Database Performance

Maji et al [285] Patch-based End-to-end Vessel segmentation DRIVE AUC = 0.9283

ensemble of ACC = 94.7
CNN
Liskowski and Patch-based CNN End-to-end Vessel segmentation DRIVE, STARE, AUC = 0.9790, 0.9928, 0.988
Krawiec [217] CHASE ACC = 95.35, 97.29, 96.96

Maninis et al [217] Fully CNN (FCN) Transfer learning Vessel segmentation DRIVE, STARE —
Wu et al [100] tracking/patch- End-to-end Vessel segmentation DRIVE AUC = 0.9701
based CNN/
PCA as
classifier
Dasgupta and Patch-based FCN End-to-end Vessel segmentation DRIVE AUC = 0.974
Singh [287] ACC = 95.33

12-28
Tan et al [216] Patch-based CNN End-to-end Vessel segmentation DRIVE, STARE ACC = 94.70, 95.45
Mo and Zhang Multi-level FCN End-to-end Vessel segmentation DRIVE, STARE, AUC = 0.9782, 0.9885, 0.9812
[289] CHASE ACC = 95.21, 96.76, 95.99

Maji et al [285] Patch-based AE Transfer learning Vessel segmentation DRIVE AUC = 0.9195, ACC = 93.27
Li et al [291] Patch-based AE End-to-end Vessel segmentation DRIVE, STARE, AUC = 0.9738, –, 0.9716,
CHASE ACC = 95.27, 96.28, 95.81
Lahiri et al [290] Patch-based AE Transfer learning Vessel segmentation DRIVE ACC = 95.3
Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Fu et al [288] Patch-based CNN End-to-end Vessel segmentation DRIVE, STARE, ACC = 95.23, 95.85, 94.89
as RNN CHASE
Lim et al [292] CNN with End-to-end ONH segmentation MESSIDOR, SEED-DB —
exaggeration
Tan et al [216] CNN End-to-end ONH segmentation DRIVE ACC = 87.90
 Sevastopolsky Modified U-Net Transfer learning ONH segmentation DRION-DB —
[294] CNN
Zilly et al [201] Multi-scale CNN End-to-end ONH segmentation DRISHTI-GS —
Alghamdi et al Cascade CNN, End-to-end ONH detection DRIVE, DIARETDB1, ACC = 100, 98.88, 99.20, 86.71
[295] each model MESSIDOR,
STARE,
Xu et al [177] CNN based on Transfer learning ONH detection MESSIDOR, STARE ACC = 99.43, 89
deconvolution
Haloi [296] 9-layer CNN End-to-end MA detection MESSIDOR, ROC AUC = 0.982
ACC = 95.4

van Grinsven et al Patches based End-to-end HE detection Kaggle, MESSIDOR AUC = 0.917, 0.979
[111] selective
sampling
Gondal et al [301] CNN model Transfer learning HE detection DIARETDB1 —
Quellec et al [276] CNN model Transfer learning HE detection DIARETDB1 AUC = 0.614

12-29
Orlando et al [114] HEF + CNN and End-to-end MA-HE detection DIARETDB1, e-ophtha, AUC = 0.8932
RF classifier MESSIDOR
Tavakoli et al [2, Patches CNN End-to-end MA detection DRIVE, DIARETDB1, ACC = 95.97
6, 16] MESSIDOR
Shan and Li [298] Patches based AE Transfer learning MA detection DIARETDB ACC = 91.38
Prentasic and 11-layer CNN, End-to-end EX detection DRiDB —
Loncaric [303]
Gondal et al [301] CNN model Transfer learning EX detection DIARETDB1
Quellec et al [276] CNN model Transfer learning EX detection DIARETDB1 AUC = 0.974,
Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

for detecting of other bright lesions such as EXs. There are a number of studies
that work on both ONH detection and segmentation using the concept of DL [201,
292–295].
Lim et al [292] presented a nine-layer CNN model for ONH segmentation. Their
algorithm involved different steps: detecting the region around the ONH, enhance-
ment, classification at pixel-level using a CNN model to predict the ONH. Tan et al
[216] identified the ONH and vessels jointly using pixel patch-based CNN. Zilly et al
[201] presented an ONH segmentation algorithm based on a multiscale two-layer
CNN model. First, the area around the ONH was cropped, and down-sampled.
After that, the area was analyzed by entropy filtering to determine the most
distinguished points and was passed to the CNN model for final segmentation.
Zhang et al [293] applied a faster CNN to localize the ONH. After identifying the
ONH, the thick vessels inside its area were eliminated by using a Hessian matrix,
and its boundaries segmented.

12.5.1.3 Microaneurysms and hemorrhages detection

Many DL-based approaches have been introduced for detection and classification of
different types of DR lesions such as MAs, and HEs [2, 6, 16, 111, 114, 274, 275,
296–300].
Haloi [296] introduced a nine-layer CNN model to group each pixel as MA or non-
MA and grading the Dr. Each pixel is categorized by taking a window around the
candidate and passing it to the CNN model. For training, they applied a data
augmentation method to create six windows around each pixel. van Grinsven et al
[111] worked on detecting HEs. Their main contribution was to overcome the over--
represented normal samples created for training a CNN model. To figure out this issue,
they employed a dynamic selective sampling idea that selects informative training
samples. Their CNN was trained using a dynamic selective sampling strategy as well.
Gondal et al [301] and Quellec et al [276] also detected HEs and small red dots.
Another similar approach was introduced by Orlando et al [4]. In their approach,
they first extracted red candidates using morphological operations. Next, they
extracted CNN and hand-engineered characteristics such as intensity and shape
from each candidate to introduce a probability map, which was utilized to make
both red lesion and image level decision. Shan and Li [298] applied the stacked
sparse AEs to identify MAs. A patch was passed to stacked sparse AEs, which
extracted characteristics, and the Softmax classifier labelled it as an MA or non-MA
patch. Tavakoli et al employed a CNN approach to detect MAs. In their approach,
they first extracted candidates using morphological functions. Next, they used CNN
classification to classify between MAs and non-MAs. In their study, the authors
compared the effect of two preprocessing methods, illumination equalization, and
top-hat transformation, on retinal images to detect MAs using a combination of
matching based approach and CNN methods [6].

12.5.1.4 Exudates detection

There are different studies on EX detections using DL [276, 301–303]. Prentasic and
Loncaric [303] employed a CNN-based algorithm for EXs detection in retinal

12-30
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

images. First, they detected the ONH, created a probability map for that. Then they
created for both vessels bright-boundary probability maps. Lastly, they applied an
11-layer CNN model to produce an EX probability map and finally segmented the
EXs. Gondal et al [301] presented an algorithm for EXs detection combined with
other DR lesions based on CNN architecture [304]. To identify DR lesions including
different types of EXs and HEs, the dense layers were removed from the CNN
model. For detecting DR lesions such as EXs at the pixel level, Quellec et al [276]
proposed a CNN visualization method. The authors did a modification on their
CNN based on the sensitivity evaluation done by Simonyan and Zisserman [305],
which helped in detecting DR and lesions by optimizing CNN predictions.

12.6 Conclusion
This chapter introduces a detailed study of methods and their results for DR
detection and classification using retinal image. DR is a complication of diabetes
that damages the retina, and causes vision loss and blindness. A robust DR
screening system will remarkably decrease the workload of clinicians. The screening
process brings sets of steps, namely identifying and segmenting the retinal land-
marks, detecting lesions, feature extraction and classification. All these steps need
different methods and techniques. Although considerable accomplishments have
been made in digital retinal image analysis, there are a variety of challenges in the
selection of suitable approaches which result in high accuracy during DR screening.

References
[1] Mookiah M R K, Acharya U R, Chua C K, Lim C M, Ng E and Laude A 2013 Computer-
aided diagnosis of diabetic retinopathy: a review Comput. Biol. Med. 43 2136–55
[2] Tavakoli M, Jazani S and Nazar M 2020 Automated detection of microaneurysms in color
fundus images using deep learning with different preprocessing approaches SPIE 11318
113180E
[3] Thomas R L, Dunstan F D, Luzio S D, Chowdhury S R, North R V and Hale S L et al
2015 Prevalence of diabetic retinopathy within a national diabetic retinopathy screening
service Br. J. Ophthalmol. 99 64–8
[4] Zana F and Klein J C 2001 Segmentation of vessel-like patterns using mathematical
morphology and curvature evaluation IEEE Trans. Image Process. 10 1010–9
[5] Nayak J, Bhat P S, Acharya R, Lim C M and Kagathi M 2008 Automated identification of
diabetic retinopathy stages using digital fundus images J. Med. Syst. 32 107–15
[6] Tavakoli M and Nazar M 2020 Comparison different vessel segmentation methods in
automated microaneurysms detection in retinal images using convolutional neural networks
SPIE 11317 113171P
[7] Cade W T 2008 Diabetes-related microvascular and macrovascular diseases in the physical
therapy setting Phys. Ther. 88 1322–35
[8] Emdin C A, Rahimi K, Neal B, Callender T, Perkovic V and Patel A 2015 Blood pressure
lowering in type 2 diabetes: a systematic review and meta-analysis JAMA 313 603–15
[9] Schaffer S W, Jong C J and Mozaffari M 2012 Role of oxidative stress in diabetes-mediated
vascular dysfunction: unifying hypothesis of diabetes revisited Vascul. Pharmacol. 57
139–49

12-31
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[10] Patz A 1980 Studies on retinal neovascularization. friedenwald lecture Investig. Ophthalmol.
Vis. Sci. 19 1133–8
[11] Niemeijer M, van Ginneken B, Russell S R, Suttorp-Schulten M S and Abramoff M D 2007
Automated detection and differentiation of drusen, exudates, and cotton-wool spots in
digital color fundus photographs for diabetic retinopathy diagnosis Investig. Ophthalmol.
Vis. Sci. 48 2260–7
[12] Tavakoli M, Mehdizadeh A, Pourreza R, Banaee T, Bahreyni Toossi M H and Pourreza H
R 2010 Early detection of diabetic retinopathy in fluorescent angiography retinal images
using image processing methods Iran. J. Med. Phys. 7 7–14
[13] Antonetti D A, Barber A J, Bronson S K, Freeman W M, Gardner T W and Jefferson L S
et al 2006 Diabetic retinopathy: seeing beyond glucose-induced microvascular disease
Diabetes 55 2401–11
[14] Watkins P J 2003 Retinopathy Brit. Med. J. 326 924–6
[15] Faust O, Acharya R, Ng E Y K, Ng K H and Suri J S 2012 Algorithms for the automated
detection of diabetic retinopathy using digital fundus images: a review J. Med. Syst. 36 145–57
[16] Tavakoli M, Nazar M and Mehdizadeh A 2020 The efficacy of microaneurysms detection
with and without vessel segmentation in color retinal images SPIE 11314 113143Y
[17] Fleming A D, Philip S, Goatman K A, Olson J A and Sharp P F 2006 Automated
assessment of diabetic retinal image quality based on clarity and field definition Investig.
Ophthalmol. Vis. Sci. 47 1120–5
[18] Pourreza H R, Bahreyni Toossi M H, Mehdizadeh A, Pourreza R and Tavakoli M 2009
Automatic detection of microaneurysms in color fundus images using a local radon
transform method Iran. J. Med. Phys. 6 13–20
[19] Tavakoli M, Shahri R P, Pourreza H, Mehdizadeh A, Banaee T and Toosi M H B 2013 A
complementary method for automated detection of microaneurysms in fluorescein angiog-
raphy fundus images to assess diabetic retinopathy Pattern Recognit. 46 2740–53
[20] Tavakoli M, Naji M, Abdollahi A and Kalantari F 2017 Attenuation correction in spect
images using attenuation map estimation with its emission data Medical Imaging 2017:
Physics of Medical Imaging vol 10132 (Bellingham, WA: International Society for Optics
and Photonics) p 101324Z
[21] Sinthanayothin C, Boyce J F, Williamson T H, Cook H L, Mensah E and Lal S et al 2002
Automated detection of diabetic retinopathy on digital fundus images Diabet. Med. 19 105–12
[22] Walter T, Massin P, Erginay A, Ordonez R, Jeulin C and Klein J C 2007 Automatic
detection of microaneurysms in color fundus images Med. Image Anal. 11 555–66
[23] Seoud L, Hurtut T, Chelbi J, Cheriet F and Langlois J P 2015 Red lesion detection using
dynamic shape features for diabetic retinopathy screening IEEE Trans. Med. Imaging 35
1116–26
[24] Group E T D R S R et al 1991 Grading diabetic retinopathy from stereoscopic color fundus
photographsan extension of the modified airlie house classification: etdrs report number 10
Ophthalmology 98 786–806
[25] Cusick M, Chew E Y, Chan C C, Kruth H S, Murphy R P and Ferris F L III 2003
Histopathology and regression of retinal hard exudates in diabetic retinopathy after
reduction of elevated serum lipid levels Ophthalmology 110 2126–33
[26] McLeod D 2005 Why cotton wool spots should not be regarded as retinal nerve fibre layer
infarcts Br. J. Ophthalmol. 89 229–37

12-32
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[27] Chui T Y, Thibos L N, Bradley A and Burns S A 2009 The mechanisms of vision loss
associated with a cotton wool spot Vis. Res. 49 2826–34
[28] Vallabha D, Dorairaj R, Namuduri K and Thompson H 2004 Automated detection and
classification of vascular abnormalities in diabetic retinopathy Conf. Record of the Thirty-
Eighth Asilomar Conf. on Signals, Systems and Computers vol 2 (Piscataway, NJ: IEEE)
pp 1625–9
[29] Williams R, Airey M, Baxter H, Forrester J, Kennedy-Martin T and Girach A 2004
Epidemiology of diabetic retinopathy and macular oedema: a systematic review Eye 18
963–83
[30] Giancardo L, Meriaudeau F, Karnowski T P, Li Y, Garg S and Tobin K W Jr et al 2012
Exudate-based diabetic macular edema detection in fundus images using publicly available
datasets Med. Image Anal. 16 216–26
[31] Taylor S R, Lightman S L, Sugar E A, Jaffe G J, Freeman W R and Altaweel M M et al
2012 The impact of macular edema on visual function in intermediate, posterior, and
panuveitis Ocular Immunol. Inflamm. 20 171–81
[32] Molina-Casado J M, Carmona E J and Garcia-Feijoo J 2017 Fast detection of the main
anatomical structures in digital retinal images based on intra-and inter-structure relational
knowledge Comput. Methods Programs Biomed. 149 55–68
[33] Moccia S, De Momi E, El Hadji S and Mattos L S 2018 Blood vessel segmentation
algorithmsreview of methods, datasets and evaluation metrics Comput. Methods Programs
Biomed. 158 71–91
[34] Wilkinson C, Ferris F L III, Klein R E, Lee P P, Agardh C D and Davis M et al 2003
Proposed international clinical diabetic retinopathy and diabetic macular edema disease
severity scales Ophthalmology 110 1677–82
[35] Group E T D R S R et al 1991 Classification of diabetic retinopathy from fluorescein
angiograms: Etdrs report number 11 Ophthalmology 98 807–22
[36] Philip S, Fleming A D, Goatman K A, Fonseca S, Mcnamee P and Scotland G S et al 2007
The efficacy of automated disease/no disease grading for diabetic retinopathy in a system-
atic screening programme Br. J. Ophthalmol. 91 1512–7
[37] Salamat N, Missen M M S and Rashid A 2019 Diabetic retinopathy techniques in retinal
images: a review Artif. Intell. Med. 97 168–88
[38] Venkatesan R, Chandakkar P, Li B and Li H K 2012 Classification of diabetic retinopathy
images using multi-class multiple-instance learning based on color correlogram features
2012 Annual Int. Conf. of the IEEE Engineering in Medicine and Biology Society
(Piscataway, NJ: IEEE) pp 1462–5
[39] Patton N, Aslam T M, MacGillivray T, Deary I J, Dhillon B and Eikelboom R H et al 2006
Retinal image analysis: concepts, applications and potential Prog. Retin. Eye Res. 25 99–127
[40] Kanski J J 2009 Clinical Ophthalmology: A Synopsis (Amsterdam: Elsevier Health Science)
[41] Zachariah S, Wykes W and Yorston D 2015 Grading diabetic retinopathy (Dr) using the
scottish grading protocol Community Eye Health 28 72
[42] Scanlon P H 2017 The english national screening programme for diabetic retinopathy 2003–
2016 Acta Diabetol. 54 515–25
[43] Ciulla T A, Amador A G and Zinman B 2003 Diabetic retinopathy and diabetic macular
edema: pathophysiology, screening, and novel therapies Diabet. Care 26 2653–64
[44] Abr’amoff M D, Garvin M K and Sonka M 2010 Retinal imaging and image analysis IEEE
Rev. Biomed. Eng. 3 169–208

12-33
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[45] Kauppi T, Kalesnykiene V, Kamarainen J K, Lensu L, Sorri I and Raninen A et al 2007

The DIARETDB1 diabetic retinopathy database and evaluation protocol Proc. of the
British Machine Conf. 1 1–10
[46] Budai A, Bock R, Maier A, Hornegger J and Michelson G 2013 Robust vessel
segmentation in fundus images Int. J. Biomed. Imaging 2013 154860
[47] Hoover A, Kouznetsova V and Goldbaum M 2000 Locating blood vessels in retinal images
by piecewise threshold probing of a matched filter response IEEE Trans. Med. Imaging 19
203–10
[48] Niemeijer M, Van Ginneken B, Cree M J, Mizutani A, Quellec G and Sánchez C I et al
2009 Retinopathy online challenge: automatic detection of microaneurysms in digital color
fundus photographs IEEE Trans. Med. Imaging 29 185–95
[49] Niemeijer M, Staal J, van Ginneken B, Loog M and Abramoff M D 2004 Comparative
study of retinal vessel segmentation methods on a new publicly available database SPIE
Medical Imaging vol 5370 (Bellingham, WA: SPIE) p 648–56
[50] Decenci’ere E, Zhang X, Cazuguel G, Lay B, Cochener B and Trone C et al 2014 Feedback
on a publicly distributed image database: the messidor database Image Anal. Stereol. 33
231–4
[51] Decenci’ere E, Cazuguel G, Zhang X, Thibault G, Klein J C and Meyer F et al 2013
Teleophta: Machine learning and image processing methods for teleophthalmology Irbm 34
196–203
[52] Fraz M M, Remagnino P, Hoppe A, Uyyanonvara B, Rudnicka A R and Owen C G et al
2012 Blood vessel segmentation methodologies in retinal images–a survey Comput. Methods
Programs Biomed. 108 407–33
[53] Pourreza-Shahri R, Tavakoli M and Kehtarnavaz N 2014 Computationally efficient optic
nerve head detection in retinal fundus images Biomed. Signal Process. Control 11 63–73
[54] Kaggle retinal database https://kaggle.com/c/diabetic-retinopathy-detection/data Accessed:
2018-1-08
[55] E-ophtha retinal database http://adcis.net/en/Download-Third-Party/E-Ophtha.html
Accessed: 2018-1-08
[56] Fraz M M, Remagnino P, Hoppe A, Uyyanonvara B, Rudnicka A R and Owen C G et al
2012 An ensemble classification-based approach applied to retinal blood vessel segmenta-
tion IEEE Trans. Biomed. Eng. 59 2538–48
[57] Aria retinal database http://eyecharity.com/aria_online.html Accessed: 2018-2-28
[58] Seed-db retinal database https://seri.com.sg/key-programmes/singapore-epidemiology-of-
eye-diseases-seed/ Accessed: 2018-8-08
[59] EyePACS retinal database http://eyepacs.com/eyepacssystem/ Accessed: 2018-3-01
[60] Zhang Z, Yin F S, Liu J, Wong W K, Tan N M and Lee B H et al 2010 Origa-light: an online
retinal fundus image database for glaucoma analysis and research 2010 Annual Int. Conf. of
the IEEE Engineering in Medicine and Biology (Piscataway, NJ: IEEE) pp 3065–8
[61] Esmaeili M, Rabbani H and Dehnavi A M 2012 Automatic optic disk boundary extraction
by the use of curvelet transform and deformable variational level set model Pattern
Recognit. 45 2832–42
[62] Qureshi R J, Kovacs L, Harangi B, Nagy B, Peto T and Hajdu A 2012 Combining
algorithms for automatic detection of optic disc and macula in fundus images Comput. Vis.
Image Understanding 116 138–45

12-34
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[63] Welfer D, Scharcanski J and Marinho D 2013 A morphologic two-stage approach for
automated optic disk detection in color eye fundus images Pattern Recognit. Lett. 34 476–85
[64] Harangi B and Hajdu A 2015 Detection of the optic disc in fundus images by combining
probability models Comput. Biol. Med. 65 10–24
[65] Akyol K, Sen B and Bayir S 2016 Automatic detection of optic disc in retinal image by
using keypoint detection, texture analysis, and visual dictionary techniques Comput. Math.
Method. Med. 2016 6814791
[66] Dai B, Wu X and Bu W 2017 Optic disc segmentation based on variational model with
multiple energies Pattern Recognit. 64 226–35
[67] Alshayeji M and Al-Roomi S A et al 2017 Optic disc detection in retinal fundus images
using gravitational law-based edge detection Med. Biol. Eng. Comput. 55 935–48
[68] Basit A and Fraz M M 2015 Optic disc detection and boundary extraction in retinal images
Appl. Opt. 54 3440–7
[69] Hsu W, Pallawala P, Lee M L and Eong K G A 2001 The Role of Domain Knowledge in
the Detection of Retinal Hard Exudates vol 2 (Piscataway, NJ: IEEE) p II
[70] Niemeijer M, Abramoff M D, Van and Ginneken B 2009 Information fusion for diabetic
retinopathy cad in digital color fundus photographs IEEE Trans. Med. Imaging 28 775–85
[71] Quellec G, Russell S R and Abra’moff M D 2010 Optimal filter framework for automated,
instantaneous detection of lesions in retinal images IEEE Trans. Med. Imaging 30 523–33
[72] Amel F, Mohammed M and Abdelhafid B 2012 Improvement of the hard exudates
detection method used for computer-aided diagnosis of diabetic retinopathy Int. J. Image,
Graph. Signal Process. 4 19–27
[73] Sánchez C I, Niemeijer M, Dumitrescu A V, Suttorp-Schulten M S, Abramoff M D and
van Ginneken B 2011 Evaluation of a computer-aided diagnosis system for diabetic
retinopathy screening on public data Investig. Ophthalmol. Vis. Sci. 52 4866–71
[74] Mansour R F 2017 Evolutionary computing enriched computer-aided diagnosis system for
diabetic retinopathy: a survey IEEE Rev. Biomed. Eng. 10 334–49
[75] Kumar D, Taylor G W and Wong A 2019 Discovery radiomics with clear-Dr:
interpretable computer aided diagnosis of diabetic retinopathy IEEE Access 7 25891–6
[76] Ganesan K, Martis R J, Acharya U R, Chua C K, Min L C and Ng E et al 2014 Computer
aided diabetic retinopathy detection using trace transforms on digital fundus images Med.
Biol. Eng. Comput. 52 663–72
[77] Sim D A, Keane P A, Tufail A, Egan C A, Aiello L P and Silva P S 2015 Automated retinal
image analysis for diabetic retinopathy in telemedicine Curr. Diabet. Rep. 15 14
[78] Mandrekar J N 2010 Receiver operating characteristic curve in diagnostic test assessment J.
Thorac. Oncol. 5 1315–6
[79] Marín D, Aquino A, Gegúndez-Arias M E and Bravo J M 2010 A new supervised method
for blood vessel segmentation in retinal images by using gray-level and moment invari-
antsbased features IEEE Trans. Med. Imaging 30 146–58
[80] Yan Z, Yang X and Cheng K T 2017 A skeletal similarity metric for quality evaluation of
retinal vessel segmentation IEEE Trans. Med. Imaging 37 1045–57
[81] Bharkad S 2017 Automatic segmentation of optic disk in retinal images Biomed. Signal
Process. Control 31 483–98
[82] Leontidis G, Al-Diri B and Hunter A 2017 A new unified framework for the early
detection of the progression to diabetic retinopathy from fundus images Comput. Biol.
Med. 90 98–115

12-35
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[83] Zou B, Chen C, Zhu C, Duan X and Chen Z 2018 Classified optic disc localization
algorithm based on verification model Comput. Graph. 70 281–7
[84] Xiong L and Li H 2016 An approach to locate optic disc in retinal images with pathological
changes Comput. Med. Imaging Graph. 47 40–50
[85] Sarathi M P, Dutta M K, Singh A and Travieso C M 2016 Blood vessel inpainting based
technique for efficient localization and segmentation of optic disc in digital fundus images
Biomed. Signal Process. Control 25 108–17
[86] Lu S 2011 Accurate and efficient optic disc detection and segmentation by a circular
transformation IEEE Trans. Med. Imaging 30 2126–33
[87] Lowell J, Hunter A, Steel D, Basu A, Ryder R and Fletcher E et al 2004 Optic nerve head
segmentation IEEE Trans. Med. Imaging 23 256–64
[88] Hoover A and Goldbaum M 2003 Locating the optic nerve in a retinal image using the
fuzzy convergence of the blood vessels IEEE Trans. Med. Imaging 22 951–8
[89] Dash J and Bhoi N 2017 A thresholding based technique to extract retinal blood vessels
from fundus images Future Comput. Inform. J. 2 103–9
[90] Zhu C, Zou B, Zhao R, Cui J, Duan X and Chen Z et al 2017 Retinal vessel segmentation
in colour fundus images using extreme learning machine Comput. Med. Imaging Graph. 55
68–77
[91] Neto L C, Ramalho G L, Neto J F R, Veras R M and Medeiros F N 2017 An unsupervised
coarseto-fine algorithm for blood vessel segmentation in fundus images Expert Syst. Appl.
78 182–92
[92] Xu L and Luo S 2010 A novel method for blood vessel detection from retinal images
Biomed. Eng. Online 9 14
[93] Nguyen U T, Bhuiyan A, Park L A and Ramamohanarao K 2013 An effective retinal blood
vessel segmentation method using multi-scale line detection Pattern Recognit. 46 703–15
[94] Zhang L, Fisher M and Wang W 2015 Retinal vessel segmentation using multi-scale textons
derived from keypoints Comput. Med. Imaging Graph. 45 47–56
[95] Fraz M M, Barman S, Remagnino P, Hoppe A, Basit A and Uyyanonvara B et al 2012 An
approach to localize the retinal blood vessels using bit planes and centerline detection
Comput. Methods Programs Biomed. 108 600–16
[96] Marín D, Aquino A, Gegúndez-Arias M E and Bravo J M 2011 A new supervised method
for blood vessel segmentation in retinal images by using gray-level and moment invari-
antsbased features IEEE Trans. Med. Imaging 30 146–58
[97] Akram M U and Khan S A 2013 Multilayered thresholding-based blood vessel segmenta-
tion for screening of diabetic retinopathy Eng. Comput. 29 165–73
[98] Vermeer K A, Vos F M, Lemij H G and Vossepoel A M 2004 A model based method for
retinal blood vessel detection Comput. Biol. Med. 34 209–19
[99] Chakraborti T, Jha D K, Chowdhury A S and Jiang X 2015 A self-adaptive matched filter
for retinal blood vessel detection Mach. Vis. Appl. 26 55–68
[100] Wu D, Zhang M, Liu J C and Bauman W 2006 On the adaptive detection of blood vessels
in retinal images IEEE Trans. Biomed. Eng. 53 341–3
[101] Tavakoli M, Mehdizadeh A, Pourreza R, Pourreza H R, Banaee T and Toosi M B 2011
Radon transform technique for linear structures detection: application to vessel detection in
fluorescein angiography fundus images Nuclear Science Symp. Conf. Record (Piscataway,
NJ: IEEE) pp 3051–6

12-36
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[102] Roychowdhury S, Koozekanani D D and Parhi K K 2015 Blood vessel segmentation of

fundus images by major vessel extraction and subimage classification IEEE J. Biomed.
Health Inform. 19 1118–28
[103] Tolias Y A and Panas S M 1998 A fuzzy vessel tracking algorithm for retinal images based
on fuzzy clustering IEEE Trans. Med. Imaging 17 263–73
[104] Azzopardi G, Strisciuglio N, Vento M and Petkov N 2015 Trainable cosfire filters for vessel
delineation with application to retinal images Med. Image Anal. 19 46–57
[105] Zana F and Klein J C 1999 A multimodal registration algorithm of eye fundus images using
vessels detection and hough transform IEEE Trans. Med. Imaging 18 419–28
[106] Soares J V, Leandro J J, Cesar R M, Jelinek H F and Cree M J 2006 Retinal vessel
segmentation using the 2-d gabor wavelet and supervised classification IEEE Trans. Med.
Imaging 25 1214–22
[107] Quellec G, Lamard M, Josselin P M, Cazuguel G, Cochener B and Roux C 2008 Optimal
wavelet transform for the detection of microaneurysms in retina photographs IEEE Trans.
Med. Imaging 27 1230–41
[108] Wang S, Tang H L, Hu Y, Sanei S, Saleh G M and Peto T et al 2016 Localizing
microaneurysms in fundus images through singular spectrum analysis IEEE Trans. Biomed.
Eng. 64 990–1002
[109] Srivastava R, Duan L, Wong D W, Liu J and Wong T Y 2017 Detecting retinal
microaneurysms and hemorrhages with robustness to the presence of blood vessels
Comput. Methods Programs Biomed. 138 83–91
[110] Dai L, Fang R, Li H, Hou X, Sheng B and Wu Q et al 2018 Clinical report guided retinal
microaneurysm detection with multi-sieving deep learning IEEE Trans. Med. Imaging 37
1149–61
[111] van Grinsven M J, van Ginneken B, Hoyng C B, Theelen T and Sánchez C I 2016 Fast
convolutional neural network training using selective data sampling: application to
hemorrhage detection in color fundus images IEEE Trans. Med. Imaging 35 1273–84
[112] Pereira C, Veiga D, Mahdjoub J, Guessoum Z, Gonçalves L and Ferreira M et al 2014
Using a multi-agent system approach for microaneurysm detection in fundus images Artif.
Intell. Med. 60 179–88
[113] Liu Q, Zou B, Chen J, Ke W, Yue K and Chen Z et al 2017 A location-to-segmentation
strategy for automatic exudate segmentation in colour retinal fundus images Comput. Med.
Imaging Graph. 55 78–86
[114] Orlando J I, Prokofyeva E, del Fresno M and Blaschko M B 2018 An ensemble deep
learning based approach for red lesion detection in fundus images Comput. Methods
Programs Biomed. 153 115–27
[115] Zhang X, Thibault G, Decenci’ere E, Marcotegui B, Laÿ B and Danno R et al 2014
Exudate detection in color retinal images for mass screening of diabetic retinopathy Med.
Image Anal. 18 1026–43
[116] Youssef D and Solouma N H 2012 Accurate detection of blood vessels improves the
detection of exudates in color fundus images Comput. Methods Programs Biomed. 108
1052–61
[117] Fraz M M, Jahangir W, Zahid S, Hamayun M M and Barman S A 2017 Multiscale
segmentation of exudates in retinal images using contextual cues and ensemble classification
Biomed. Signal Process. Control 35 50–62

12-37
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[118] Amin J, Sharif M, Yasmin M, Ali H and Fernandes S L 2017 A method for the detection
and classification of diabetic retinopathy using structural predictors of bright lesions J.
Comput. Sci. 19 153–64
[119] Sopharak A, Uyyanonvara B and Barman S 2009 Automatic exudate detection for diabetic
retinopathy screening Sci. Asia 35 80–8
[120] Sánchez C I, García M, Mayo A, Lo�ez M I and Hornero R 2009 Retinal image analysis
based on mixture models to detect hard exudates Med. Image Anal. 13 650–8
[121] Koh J E, Acharya U R, Hagiwara Y, Raghavendra U, Tan J H and Sree S V et al 2017
Diagnosis of retinal health in digital fundus images using continuous wavelet transform
(cwt) and entropies Comput. Biol. Med. 84 89–97
[122] Gardner G, Keating D, Williamson T and Elliott A 1996 Automatic detection of diabetic
retinopathy using an artificial neural network: a screening tool Br. J. Ophthalmol. 80 940–4
[123] Rubini S S and Kunthavai A 2015 Diabetic retinopathy detection based on eigenvalues of
the hessian matrix Procedia Comput. Sci. 47 311–8
[124] Gulshan V, Peng L, Coram M, Stumpe M C, Wu D and Narayanaswamy A et al 2016
Development and validation of a deep learning algorithm for detection of diabetic
retinopathy in retinal fundus photographs JAMA 316 2402–10
[125] Köse C, Şevik U, IKibaş C and Erdöl H 2012 Simple methods for segmentation and
measure-˙ ment of diabetic retinopathy lesions in retinal fundus images Comput. Methods
Programs Biomed. 107 274–93
[126] Yang Y, Li T, Li W, Wu H, Fan W and Zhang W 2017 Lesion detection and grading of
diabetic retinopathy via two-stages deep convolutional neural networks Int. Conf.
on Medical Image Computing and Computer-Assisted Intervention (Berlin: Springer)
pp 533–40
[127] Sisodia D S, Nair S and Khobragade P 2017 Diabetic retinal fundus images: preprocessing
and feature extraction for early detection of diabetic retinopathy Biomed. Pharmacol. J. 10
615–26
[128] Kumar P S, Deepak R, Sathar A, Sahasranamam V and Kumar R R 2016 Automated
detection system for diabetic retinopathy using two field fundus photography Procedia
Comput. Sci. 93 486–94
[129] Gupta G, Kulasekaran S, Ram K, Joshi N, Sivaprakasam M and Gandhi R 2017 Local
characterization of neovascularization and identification of proliferative diabetic retinop-
athy in retinal fundus images Comput. Med. Imaging Graph. 55 124–32
[130] Sinthanayothin C, Kongbunkiat V, Phoojaruenchanachai S and Singalavanija A 2003
Automated screening system for diabetic retinopathy 3rd Int. Symp. on Image and Signal
Processing and Analysis, 2003. ISPA 2003. Proc. of the vol 2 (Piscataway, NJ: IEEE) pp
915–20
[131] Usher D, Dumskyj M, Himaga M, Williamson T H, Nussey S and Boyce J 2004
Automated detection of diabetic retinopathy in digital retinal images: a tool for diabetic
retinopathy screening Diabet. Med. 21 84–90
[132] Fleming A, Goatman K, Williams G, Philip S, Sharp P and Olson J 2008 Automated
detection of blot haemorrhages as a sign of referable diabetic retinopathy Proc. Medical
Image Understanding and Analysis
[133] Perumalsamy N, Prasad N M, Sathya S and Ramasamy K 2007 Software for reading and
grading diabetic retinopathy: aravind diabetic retinopathy screening 3.0 Diabet. Care 30
2302–6

12-38
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[134] Winder R J, Morrow P J, McRitchie I N, Bailie J and Hart P M 2009 Algorithms for digital
image processing in diabetic retinopathy Comput. Med. Imaging Graph. 33 608–22
[135] Mookiah M R K, Acharya U R, Martis R J, Chua C K, Lim C M and Ng E et al 2013
Evolutionary algorithm based classifier parameter tuning for automatic diabetic retinop-
athy grading: A hybrid feature extraction approach Knowl. Based Syst. 39 9–22
[136] Abramoff M D and Suttorp-Schulten M S 2005 Web-based screening for diabetic
retinopathy in a primary care population: the eyecheck project Telemed. J. e-Health 11
668–74
[137] Abrámoff M D, Niemeijer M, Suttorp-Schulten M S, Viergever M A, Russell S R and Van
Ginneken B 2008 Evaluation of a system for automatic detection of diabetic retinopathy
from color fundus photographs in a large population of patients with diabetes Diabet. Care
31 193–8
[138] Teng T, Lefley M and Claremont D 2002 Progress towards automated diabetic ocular
screening: a review of image analysis and intelligent systems for diabetic retinopathy Med.
Biol. Eng. Comput. 40 2–13
[139] Wang H, Hsu W, Goh K G and Lee M L 2000 An effective approach to detect lesions in
color retinal images Proc. IEEE Conf. on Computer Vision and Pattern Recognition vol 2
(Piscataway, NJ: IEEE) pp 181–6
[140] Singalavanija A, Supokavej J, Bamroongsuk P, Sinthanayothin C, Phoojaruenchanachai S
and Kongbunkiat V 2006 Feasibility study on computer-aided screening for diabetic
retinopathy Jpn. J. Ophthalmol. 50 361–6
[141] Gang L, Chutatape O and Krishnan S M 2002 Detection and measurement of retinal
vessels in fundus images using amplitude modified second-order gaussian filter IEEE Trans.
Biomed. Eng. 49 168–72
[142] Kahai P, Namuduri K R and Thompson H 2006 A decision support framework for
automated screening of diabetic retinopathy Int. J. Biomed. Imaging 2006 045806
[143] Yun W L, Acharya U R, Venkatesh Y V, Chee C, Min L C and Ng E Y K 2008
Identification of different stages of diabetic retinopathy using retinal optical images Inf. Sci.
178 106–21
[144] Acharya U R, Lim C M, Ng E Y K, Chee C and Tamura T 2009 Computer-based detection
of diabetes retinopathy stages using digital fundus images Proc. Inst. Mech. Eng. H 223
545–53
[145] Larsen N, Godt J, Grunkin M, Lund-Andersen H and Larsen M 2003 Automated detection
of diabetic retinopathy in a fundus photographic screening population Investig. Ophthalmol.
Vis. Sci. 44 767–71
[146] Joshi G D, Sivaswamy J and Krishnadas S 2011 Optic disk and cup segmentation from
monocular color retinal images for glaucoma assessment IEEE Trans. Med. Imaging 30
1192–205
[147] Tavakoli M, Kelley P, Nazar M and Kalantari F 2017 Automated fovea detection based on
unsupervised retinal vessel segmentation method 2017 IEEE Nuclear Science Symp. and
Medical Imaging Conf. (NSS/MIC) (Piscataway, NJ: IEEE) pp 1–7
[148] Sinthanayothin C 1999 Image analysis for automatic diagnosis of diabetic retinopathy PhD
Thesis (Department of Physics, King’s College London)
[149] Gagnon L, Lalonde M, Beaulieu M and Boucher M C 2001 Procedure to detect anatomical
structures in optical fundus images Medical Imaging 2001: Image Processing vol 4322
(Bellingham, WA: IEEE) pp 1218–25

12-39
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[150] Hubbard L D, Brothers R J, King W N, Clegg L X, Klein R and Cooper L S et al 1999

Methods for evaluation of retinal microvascular abnormalities associated with hypertension/
sclerosis in the atherosclerosis risk in communities study Ophthalmology 106 2269–80
[151] Kamble R, Kokare M, Deshmukh G, Hussin F A and Mériaudeau F 2017 Localization of
optic disc and fovea in retinal images using intensity based line scanning analysis Comput.
Biol. Med. 87 382–96
[152] Elbalaoui A, Fakir M, Boutaounte M and Merbouha A 2017 Automatic localization of the
optic disc center in retinal images based on angle detection in curvature scale space Medical
Imaging: Concepts, Methodologies, Tools, and Applications. (Hershey, PA: IGI Global) pp
679–92
[153] Unver H M, Kökver Y, Duman E and Erdem O A 2019 Statistical edge detection and
circular� hough transform for optic disk localization Appl.Sci. 9 350
[154] Meng X, Xi X, Yang L, Zhang G, Yin Y and Chen X 2018 Fast and effective optic disk
localization based on convolutional neural network Neurocomputing 312 285–95
[155] Muangnak N, Aimmanee P and Makhanov S 2018 Automatic optic disk detection in retinal
images using hybrid vessel phase portrait analysis Med. Biol. Eng. Comput. 56 583–98
[156] Lalonde M, Gagnon L and Boucher M C 2000 Non-recursive paired tracking for vessel
extraction from retinal images Vision Interface. pp 61–8
[157] Zahoor M N and Fraz M M 2017 Fast optic disc segmentation in retina using polar
transform IEEE Access 5 12293–300
[158] Manju K, Sabeenian R and Surendar A 2017 A review on optic disc and cup segmentation
Biomed. Pharmacol. J. 10 373–9
[159] Rodrigues L C and Marengoni M 2017 Segmentation of optic disc and blood vessels in
retinal images using wavelets, mathematical morphology and hessian-based multi-scale
filtering Biomed. Signal Process. Control 36 39–49
[160] Rehman Z U, Naqvi S S, Khan T M, Arsalan M, Khan M A and Khalil M 2019 Multi-
parametric optic disc segmentation using superpixel based feature classification Expert Syst.
Appl. 120 461–73
[161] Goldbaum M, Moezzi S, Taylor A, Chatterjee S, Boyd J and Hunter E et al 1996
Automated diagnosis and image understanding with object extraction, object classification,
and inferencing in retinal images Proc. of 3rd Int. Conf. on Image Processing vol 3
(Piscataway, NJ: IEEE) pp 695–8
[162] Sinthanayothin C, Boyce J F, Cook H L and Williamson T H 1999 Automated localisation
of the optic disc, fovea, and retinal blood vessels from digital colour fundus images Br. J.
Ophthalmol. 83 902–10
[163] Walter T and Klein J C 2001 Segmentation of color fundus images of the human retina:
Detection of the optic disc and the vascular tree using morphological techniques Int. Symp.
on Medical Data Analysis. (Berlin: Springer) pp 282–7
[164] Tavakoli M, Nazar M, Golestaneh A and Kalantari F 2017 Automated optic nerve head
detection based on different retinal vasculature segmentation methods and mathematical
morphology Nuclear Science Symp. and Medical Imaging Conf. (NSS/MIC) (Piscataway,
NJ: IEEE) pp 1–7
[165] Chrástek R, Wolf M, Donath K, Niemann H, Paulus D and Hothorn T et al 2005
Automated segmentation of the optic nerve head for diagnosis of glaucoma Med. Image
Anal. 9 297–314

12-40
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[166] Abdel-Ghafar R and Morris T 2007 Progress towards automated detection and character-
ization of the optic disc in glaucoma and diabetic retinopathy Medical Inform. Internet
Med. 32 19–25
[167] Tavakoli M, Toosi M B, Pourreza R, Banaee T and Pourreza H R 2011 Automated optic
nerve head detection in fluorescein angiography fundus images Nuclear Science Symp.
Conf. Record (Piscataway, NJ: IEEE) pp 3057–60
[168] Sánchez C I, Hornero R, López M I and Poza J 2004 Retinal image analysis to detect and
quantify lesions associated with diabetic retinopathy The 26th Annual Int. Conf. of the
Engineering in Medicine and Biology Society vol 1 (Piscataway, NJ: IEEE) pp 1624–7
[169] Li H and Chutatape O 2004 Automated feature extraction in color retinal images by a
model based approach IEEE Trans. Biomed. Eng. 51 246–54
[170] Foracchia M, Grisan E and Ruggeri A 2004 Detection of optic disc in retinal images by
means of a geometrical model of vessel structure IEEE Trans. Med. Imaging 23 1189–95
[171] Youssif A A H A R, Ghalwash A Z and Ghoneim A A S A R 2007 Optic disc detection
from normalized digital fundus images by means of a vessels’ direction matched filter IEEE
Trans. Med. Imaging 27 11–8
[172] Lu S and Lim J H 2010 Automatic optic disc detection from retinal images by a line
operator IEEE Trans. Biomed. Eng. 58 88–94
[173] Li H and Chutatape O 2001 Automatic location of optic disk in retinal images Proc. 2001
Int. Conf. on Image Processing (Cat. No. 01CH37205) vol 2 (Piscataway, NJ: IEEE) pp
837–40
[174] Li H and Chutatape O 2003 A model-based approach for automated feature extraction in
fundus images Innull (Piscataway, NJ: IEEE) 394
[175] Osareh A, Mirmehdi M, Thomas B and Markham R 2002 Comparison of colour spaces for
optic disc localisation in retinal images Object Recognition Supported by User Interaction for
Service Robots vol 1 (Piscataway, NJ: IEEE) pp 743–6
[176] Osareh A, Mirmehdi M, Thomas B and Markham R 2002 Classification and localisation of
diabetic-related eye disease European Conference on Computer Vision. (Berlin: Springer) pp
502–16
[177] Xu J, Chutatape O, Sung E, Zheng C and Kuan P C T 2007 Optic disk feature extraction via
modified deformable model technique for glaucoma analysis Pattern Recognit. 40 2063–76
[178] Wong D, Liu J, Lim J, Jia X, Yin F and Li H et al 2008 Level-set based automatic cup-to-
disc ratio determination using retinal fundus images in argali 2008 30th Annual Int. Conf.
of the IEEE Engineering in Medicine and Biology Society (Piscataway, NJ: IEEE) pp
2266–9
[179] Lalonde M, Beaulieu M and Gagnon L 2001 Fast and robust optic disc detection using
pyramidal decomposition and hausdorff-based template matching IEEE Trans. Med.
Imaging 20 1193–200
[180] Ter Haar F 2005 Automatic localization of the optic disc in digital colour images of the
human retina. (Utrecht: Utrecht University)
[181] Tobin K W, Chaum E, Govindasamy V P and Karnowski T P 2007 Detection of anatomic
structures in human retinal imagery IEEE Trans. Med. Imaging 26 1729–39
[182] Niemeijer M, Abràmoff M D and Van Ginneken B 2009 Fast detection of the optic disc
and fovea in color fundus photographs Med. Image Anal. 13 859–70
[183] Fleming A D, Goatman K A, Philip S, Olson J A and Sharp P F 2006 Automatic detection
of retinal anatomy to assist diabetic retinopathy screening Phys. Med. Biol. 52 331

12-41
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[184] Niemeijer M, Abràmoff M D and Van Ginneken B 2006 Segmentation of the optic disc,
macula and vascular arch in fundus photographs IEEE Trans. Med. Imaging 26 116–27
[185] Abramoff M D and Niemeijer M 2006 The automatic detection of the optic disc location in
retinal images using optic disc location regression 2006 International Conference of the
IEEE Engineering in Medicine and Biology Society. (Piscataway, NJ: IEEE) pp 4432–5
[186] Abramoff M D, Alward W L, Greenlee E C, Shuba L, Kim C Y and Fingert J H et al 2007
Automated segmentation of the optic disc from stereo color photographs using physiolog-
ically plausible features Investigative Ophthalmol. Vis. Sci. 48 1665–73
[187] Hsiao H K, Liu C C, Yu C Y, Kuo S W and Yu S S 2012 A novel optic disc detection
scheme on retinal images Expert Syst. Appl. 39 10600–6
[188] Yu H, Barriga E S, Agurto C, Echegaray S, Pattichis M S and Bauman W et al 2012 Fast
localization and segmentation of optic disk in retinal images using directional matched
filtering and level sets IEEE Trans. Inf. Technol. Biomed. 16 644–57
[189] Mahfouz A E and Fahmy A S 2010 Fast localization of the optic disc using projection of
image features IEEE Trans. Image Process. 19 3285–9
[190] Estrada R, Allingham M J, Mettu P S, Cousins S W, Tomasi C and Farsiu S 2015 Retinal
arteryvein classification via topology estimation IEEE Trans. Med. Imaging 34 2518–34
[191] Dashtbozorg B, Mendonça A M and Campilho A 2014 An automatic graph-based
approach for artery/vein classification in retinal images IEEE Trans. Image Process. 23
1073–83
[192] Zhao Y Q, Wang X H, Wang X F and Shih F Y 2014 Retinal vessels segmentation based
on level set and region growing Pattern Recognit. 47 2437–46
[193] Kovács G and Hajdu A 2016 A self-calibrating approach for the segmentation of retinal
vessels by template matching and contour reconstruction Med. Image Anal. 29 24–46
[194] Lázár I and Hajdu A 2015 Segmentation of retinal vessels by means of directional response
vector similarity and region growing Comput. Biol. Med. 66 209–21
[195] Orlando J I, Prokofyeva E and Blaschko M B 2017 A discriminatively trained fully
connected conditional random field model for blood vessel segmentation in fundus images
IEEE Trans. Biomed. Eng. 64 16–27
[196] Hassan M, Amin M, Murtza I, Khan A and Chaudhry A 2017 Robust hidden markov
model based intelligent blood vessel detection of fundus images Comput. Methods Programs
Biomed. 151 193–201
[197] Aslani S and Sarnel H 2016 A new supervised retinal vessel segmentation method based on
robust hybrid features Biomed. Signal Process. Control 30 1–12
[198] Lupascu C A, Tegolo D and Trucco E 2010 Fabc: retinal vessel segmentation using
adaboost IEEE Trans. Inf. Technol. Biomed. 14 1267–74
[199] Vega R, Sanchez-Ante G, Falcon-Morales L E, Sossa H and Guevara E 2015 Retinal vessel
extraction using lattice neural networks with dendritic processing Comput. Biol. Med. 58
20–30
[200] Waheed A, Akram M U, Khalid S, Waheed Z, Khan M A and Shaukat A 2015 Hybrid
features and mediods classification based robust segmentation of blood vessels J. Med. Syst.
39 128
[201] Zilly J, Buhmann J M and Mahapatra D 2017 Glaucoma detection using entropy sampling
and ensemble learning for automatic optic cup and disc segmentation Comput. Med.
Imaging Graph. 55 28–41

12-42
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[202] Annunziata R, Garzelli A, Ballerini L, Mecocci A and Trucco E 2016 Leveraging

multiscale hessian-based enhancement with a novel exudate inpainting technique for retinal
vessel segmentation IEEE J. Biomed. Health Inform. 20 1129–38
[203] Mendonca A M and Campilho A 2006 Segmentation of retinal blood vessels by combining
the detection of centerlines and morphological reconstruction IEEE Trans. Med. Imaging
25 1200–13
[204] Martinez-Perez M E, Hughes A D, Thom S A, Bharath A A and Parker K H 2007
Segmentation of blood vessels from red-free and fluorescein retinal images Med. Image
Anal. 11 47–61
[205] Zhang J, Dashtbozorg B, Bekkers E, Pluim J P, Duits R and ter Haar Romeny B M 2016
Robust retinal vessel segmentation via locally adaptive derivative frames in orientation
scores IEEE Trans. Med. Imaging 35 2631–44
[206] Lam B S Y and Yan H 2008 A novel vessel segmentation algorithm for pathological retina
images based on the divergence of vector fields IEEE Trans. Med. Imaging 27 237–46
[207] Ricci E and Perfetti R 2007 Retinal blood vessel segmentation using line operators and
support vector classification IEEE Trans. Med. Imaging 26 1357–65
[208] Wang Y and Lee S C 1997 A fast method for automated detection of blood vessels in retinal
images Conf. Record of the Thirty-First Asilomar Conf. on Signals, Systems and Computers,
1997 vol 2 (Piscataway, NJ: IEEE) pp 1700–4
[209] Jiang X and Mojon D 2003 Adaptive local thresholding by verification-based multi-
threshold probing with application to vessel detection in retinal images IEEE Trans. Pattern
Anal. Mach. Intell. 25 131–7
[210] Heneghan C, Flynn J, OKeefe M and Cahill M 2002 Characterization of changes in blood
vessel width and tortuosity in retinopathy of prematurity using image analysis Med. Image
Anal. 6 407–29
[211] Roychowdhury S, Koozekanani D D and Parhi K K 2015 Iterative vessel segmentation of
fundus images IEEE Trans. Biomed. Eng. 62 1738–49
[212] Staal J, Abra’moff M D, Niemeijer M, Viergever M A and Van Ginneken B 2004 Ridge-
based vessel segmentation in color images of the retina IEEE Trans. Med. Imaging 23 501–9
[213] Li Q, Feng B, Xie L, Liang P, Zhang H and Wang T 2016 A cross-modality learning
approach for vessel segmentation in retinal images IEEE Trans. Med. Imaging 35 109–18
[214] Annunziata R and Trucco E 2016 Accelerating convolutional sparse coding for curvi-
linear structures segmentation by refining scird-ts filter banks IEEE Trans. Med. Imaging
35 2381–92
[215] Wang S, Yin Y, Cao G, Wei B, Zheng Y and Yang G 2015 Hierarchical retinal blood vessel
segmentation based on feature and ensemble learning Neurocomputing 149 708–17
[216] Tan J H, Acharya U R, Bhandary S V, Chua K C and Sivaprasad S 2017 Segmentation of
optic disc, fovea and retinal vasculature using a single convolutional neural network J.
Comput. Sci. 20 70–9
[217] Liskowski P and Krawiec K 2016 Segmenting retinal blood vessels with deep neural
networks IEEE Trans. Med. Imaging 35 2369–80
[218] Yue K, Zou B, Chen Z and Liu Q 2019 Retinal vessel segmentation using dense u-net with
multiscale inputs J. Med. Imaging 6 034004
[219] Lin Y, Zhang H and Hu G 2018 Automatic retinal vessel segmentation via deeply
supervised and smoothly regularized network IEEE Access 7 57717–24

12-43
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[220] Ren X, Zheng Y, Zhao Y, Luo C, Wang H and Lian J et al 2018 Drusen segmentation from
retinal images via supervised feature learning IEEE Access 6 2952–61
[221] Oliveira A, Pereira S and Silva C A 2018 Retinal vessel segmentation based on fully
convolutional neural networks Expert Syst. Appl. 112 229–42
[222] Singh N, Kaur L and Singh K 2019 Segmentation of retinal blood vessels based on
featureoriented dictionary learning and sparse coding using ensemble classification
approach J. Med. Imaging 6 044006
[223] Narasimha-Iyer H, Can A, Roysam B, Tanenbaum H L and Majerovics A 2007 Integrated
analysis of vascular and nonvascular changes from color retinal fundus image sequences
IEEE Trans. Biomed. Eng. 54 1436–45
[224] Goatman K, Charnley A, Webster L and Nussey S 2011 Assessment of automated disease
detection in diabetic retinopathy screening using two-field photography PLoS One 6 e27524
[225] Köse C, Şevik U and Gençalioğlu O 2008 Automatic segmentation of age-related macular
degeneration in retinal fundus images Comput. Biol. Med. 38 611–9
[226] Zhou W, Wu C, Chen D, Yi Y and Du W 2017 Automatic microaneurysm detection using
the sparse principal component analysis-based unsupervised classification method IEEE
Access 5 2563–72
[227] Al-Bander B, Al-Nuaimy W, Williams B M and Zheng Y 2018 Multiscale sequential
convolutional neural networks for simultaneous detection of fovea and optic disc Biomed.
Signal Process. Control 40 91–101
[228] GeethaRamani R and Balasubramanian L 2018 Macula segmentation and fovea local-
ization employing image processing and heuristic based clustering for automated retinal
screening Comput. Methods Programs Biomed. 160 153–63
[229] Fadzil M A, Izhar L I and Nugroho H A 2010 Determination of foveal avascular zone in
diabetic retinopathy digital fundus images Comput. Biol. Med. 40 657–64
[230] Acharya R, Ng Y E and Suri J S 2008 Image Modeling of the Human Eye. (Boston, MA:
Artech House)
[231] Sekhar S, Al-Nuaimy W and Nandi A K 2008 Automated localisation of retinal optic disk
using Hough transform 2008 5th Int. Symp. on Biomedical Imaging: From Nano to Macro
(Piscataway, NJ: IEEE) pp 1577–80
[232] Welfer D, Scharcanski J and Marinho D 2011 Fovea center detection based on the retina
anatomy and mathematical morphology Comput. Methods Programs Biomed. 104 397–409
[233] Niemeijer M, Van Ginneken B, Staal J, Suttorp-Schulten M S and Abràmoff M D 2005
Automatic detection of red lesions in digital color fundus photographs IEEE Trans. Med.
Imaging 24 584–92
[234] Saleh M D and Eswaran C 2012 An automated decision-support system for non-
proliferative diabetic retinopathy disease based on mas and has detection Comput.
Methods Programs Biomed. 108 186–96
[235] Lazar I and Hajdu A 2013 Retinal microaneurysm detection through local rotating cross-
section profile analysis IEEE Trans. Med. Imaging 32 400–7
[236] Dashtbozorg B, Zhang J, Huang F and ter Haar Romeny B M 2018 Retinal micro-
aneurysms detection using local convergence index features IEEE Trans. Image Process. 27
3300–15
[237] Habib M, Welikala R, Hoppe A, Owen C, Rudnicka A and Barman S 2017 Detection
of microaneurysms in retinal images using an ensemble classifier Inform. Med. Unlocked 9
44–57

12-44
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[238] Gegundez-Arias M E, Marin D, Ponte B, Alvarez F, Garrido J and Ortega C et al 2017 A

tool for automated diabetic retinopathy pre-screening based on retinal image computer
analysis Comput. Biol. Med. 88 100–9
[239] Ram K, Joshi G D and Sivaswamy J 2010 A successive clutter-rejection-based approach for
early detection of diabetic retinopathy IEEE Trans. Biomed. Eng. 58 664–73
[240] Mizutani A, Muramatsu C, Hatanaka Y, Suemori S, Hara T and Fujita H 2009 Automated
microaneurysm detection method based on double ring filter in retinal fundus images
Medical Imaging 2009: Computer-Aided Diagnosis vol 7260 (Bellingham, WA:
International Society for Optics and Photonics) p 72601N
[241] Ravishankar S, Jain A and Mittal A 2009 Automated feature extraction for early detection
of diabetic retinopathy in fundus images 2009 IEEE Conf. on Computer Vision and Pattern
Recognition (Piscataway, NJ: IEEE) pp 210–7
[242] Shah S A A, Laude A, Faye I and Tang T B 2016 Automated microaneurysm detection in
diabetic retinopathy using curvelet transform J. Biomed. Opt. 21 101404
[243] Cao W, Czarnek N, Shan J and Li L 2018 Microaneurysm detection using principal
component analysis and machine learning methods IEEE Trans. Nanobiosci. 17 191–8
[244] Manjaramkar A and Kokare M 2018 Statistical geometrical features for microaneurysm
detection J. Digit. Imaging 31 224–34
[245] Mumtaz R, Hussain M, Sarwar S, Khan K, Mumtaz S and Mumtaz M 2018 Automatic
detection of retinal hemorrhages by exploiting image processing techniques for screening
retinal diseases in diabetic patients Int. J. Diabet. Dev. Countries 38 80–7
[246] Biran A, Bidari P S and Raahemifar K 2016 Automatic method for exudates and
hemorrhages detection from fundus retinal images Int. J. Comput. Inform. Eng. 10 1599–602
[247] Lahmiri S and Shmuel A 2017 Variational mode decomposition based approach for
accurate classification of color fundus images with hemorrhages Opt. Laser Technol. 96
243–8
[248] Hatanaka Y, Nakagawa T, Hayashi Y, Kakogawa M, Sawada A and Kawase K et al 2008
Improvement of automatic hemorrhage detection methods using brightness correction on
fundus images Medical Imaging 2008: Computer-Aided Diagnosis vol 6915 (Bellingham,
WA: International Society for Optics and Photonics) p 69153E
[249] Spencer T, Olson J A, McHardy K C, Sharp P F and Forrester J V 1996 An image-
processing strategy for the segmentation and quantification of microaneurysms in fluo-
rescein angiograms of the ocular fundus Comput. Biomed. Res. 29 284–302
[250] Frame A J, Undrill P E, Cree M J, Olson J A, McHardy K C and Sharp P F et al 1998 A
comparison of computer based classification methods applied to the detection of micro-
aneurysms in ophthalmic fluorescein angiograms Comput. Biol. Med. 28 225–38
[251] Zhang B, Wu X, You J, Li Q and Karray F 2010 Detection of microaneurysms using multi-
scale correlation coefficients Pattern Recognit. 43 2237–48
[252] Zhang X and Chutatape O 2005 A svm approach for detection of hemorrhages in
background diabetic retinopathy Proc. 2005 IEEE Int. Joint Conf. on Neural Networks
vol 4 (Piscataway, NJ: IEEE) pp 2435–40
[253] Hipwell J, Strachan F, Olson J, McHardy K, Sharp P and Forrester J 2000 Automated
detection of microaneurysms in digital red-free photographs: a diabetic retinopathy
screening tool Diabet. Med. 17 588–94
[254] Antal B and Hajdu A 2012 An ensemble-based system for microaneurysm detection and
diabetic retinopathy grading IEEE Trans. Biomed. Eng. 59 1720–6

12-45
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[255] Kusakunniran W, Wu Q, Ritthipravat P and Zhang J 2018 Hard exudates segmentation

based on learned initial seeds and iterative graph cut Comput. Methods Programs Biomed.
158 173–83
[256] Zubair M, Ali H and Javed M Y 2016 Automated segmentation of hard exudates using
dynamic thresholding to detect diabetic retinopathy in retinal photographs JMPT 7 109–16
[257] Lara Rodríguez L D and Urcid Serrano G 2016 Exudates and blood vessel segmentation in
eye fundus images using the fourier and cosine discrete transforms Comput. Sist. 20 697–708
[258] Osareh A, Mirmehdi M, Thomas B and Markham R 2003 Automated identification of
diabetic retinal exudates in digital colour images Br. J. Ophthalmol. 87 1220–3
[259] Osareh A, Shadgar B and Markham R 2009 A computational-intelligence-based approach
for detection of exudates in diabetic retinopathy images IEEE Trans. Inf. Technol. Biomed.
13 535–45
[260] Bezdek J C 2013 Pattern Recognition with Fuzzy Objective Function Algorithms. (Berlin:
Springer Science & Business Media)
[261] Sopharak A, Nwe K T, Moe Y A, Dailey M N and Uyyanonvara B et al 2008 Automatic
exudate detection with a naive Bayes classifier Int. Conf. on Embedded Systems and
Intelligent Technology. pp 139–42
[262] Sopharak A, Uyyanonvara B, Barman S and Williamson T H 2008 Automatic detection of
diabetic retinopathy exudates from non-dilated retinal images using mathematical mor-
phology methods Comput. Med. Imaging Graph. 32 720–7
[263] Walter T, Klein J C, Massin P and Erginay A 2002 A contribution of image processing to
the diagnosis of diabetic retinopathy-detection of exudates in color fundus images of the
human retina IEEE Trans. Med. Imaging 21 1236–43
[264] Welfer D, Scharcanski J and Marinho D 2010 A coarse-to-fine strategy for automatically
detecting exudates in color eye fundus images Comput. Med. Imaging Graph. 34 228–35
[265] Sánchez C I, Hornero R, López M I, Aboy M, Poza J and Abásolo D 2008 A novel
automatic image processing algorithm for detection of hard exudates based on retinal image
analysis Med. Eng. Phys. 30 350–7
[266] Obermeyer Z and Lee T H 2017 Lost in thought: The limits of the human mind and the
future of medicine New Engl. J. Med. 377 1209
[267] Jiang F, Jiang Y, Zhi H, Dong Y, Li H and Ma S et al 2017 Artificial intelligence in
healthcare: past, present and future Stroke Vasc. Neurol. 2 230–43
[268] Schmidt-Erfurth U, Sadeghipour A, Gerendas B S, Waldstein S M and Bogunovi´c H 2018
Artificial intelligence in retina Prog. Retin. Eye Res. 67 1–29
[269] Joshi S and Karule P 2018 A review on exudates detection methods for diabetic retinopathy
Biomed. Pharmacother. 97 1454–60
[270] Almotiri J, Elleithy K and Elleithy A 2018 Retinal vessels segmentation techniques and
algorithms: a survey Appl. Sci. 8 155
[271] Almazroa A, Burman R, Raahemifar K and Lakshminarayanan V 2015 Optic disc and
optic cup segmentation methodologies for glaucoma image detection: a survey J.
Ophthalmol. 2015 180972
[272] Thakur N and Juneja M 2018 Survey on segmentation and classification approaches of
optic cup and optic disc for diagnosis of glaucoma Biomed. Signal Process. Control 42
162–89
[273] Litjens G, Kooi T, Bejnordi B E, Setio A A A, Ciompi F and Ghafoorian M et al 2017 A
survey on deep learning in medical image analysis Med. Image Anal. 42 60–88

12-46
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

[274] Khojasteh P, Aliahmad B and Kumar D K 2018 Fundus images analysis using deep
features for detection of exudates, hemorrhages and microaneurysms BMC Ophthalmol. 18
1–13
[275] Parmar R, Lakshmanan R, Purushotham S and Soundrapandiyan R 2019 Detecting
diabetic retinopathy from retinal images using CUDA deep neural network Intelligent
Pervasive Computing Systems for Smarter Healthcare (New York: Wiley) pp 379–96
[276] Quellec G, Charrière K, Boudi Y, Cochener B and Lamard M 2017 Deep image mining for
diabetic retinopathy screening Med. Image Anal. 39 178–93
[277] Zhou L, Zhao Y, Yang J, Yu Q and Xu X 2017 Deep multiple instance learning for
automatic detection of diabetic retinopathy in retinal images IET Image Proc. 12 563–71
[278] Gargeya R and Leng T 2017 Automated identification of diabetic retinopathy using deep
learning Ophthalmology 124 962–9
[279] Son J, Shin J Y, Kim H D, Jung K H, Park K H and Park S J 2020 Development and
validation of deep learning models for screening multiple abnormal findings in retinal
fundus images Ophthalmology 127 85–94
[280] Kathiresan S, Sait A R W, Gupta D, Lakshmanaprabu S, Khanna A and Pandey H M
2020 Automated detection and classification of fundus diabetic retinopathy images using
synergic deep learning model Pattern Recognit. Lett. 133 210–6
[281] Abràmoff M D, Lou Y, Erginay A, Clarida W, Amelon R and Folk J C et al 2016
Improved automated detection of diabetic retinopathy on a publicly available dataset
through integration of deep learning Investig. Ophthalmol. Vis. Sci. 57 5200–6
[282] Wang Z, Yin Y, Shi J, Fang W, Li H and Wang X 2017 Zoom-in-net: Deep mining lesions
for diabetic retinopathy detection Int. Conf. on Medical Image Computing and Computer-
Assisted Intervention (Berlin: Springer) pp 267–75
[283] Lam C, Yu C, Huang L and Rubin D 2018 Retinal lesion detection with deep learning
using image patches Investig. Ophthalmol. Vis. Sci. 59 590–6
[284] Sayres R, Taly A, Rahimy E, Blumer K, Coz D and Hammel N et al 2019 Using a deep
learning algorithm and integrated gradients explanation to assist grading for diabetic
retinopathy Ophthalmology 126 552–64
[285] Maji D, Santara A, Mitra P and Sheet D 2016 Ensemble of deep convolutional neural
networks for learning to detect retinal vessels in fundus images. arXiv preprint
arXiv:160304833
[286] Maninis K K, Pont-Tuset J, Arbeláez P and Van Gool L 2016 Deep retinal image
understanding Int. Conf. on Medical Image Computing and Computer-Assisted Intervention
(Berlin: Springer) pp 140–8
[287] Dasgupta A and Singh S 2017 A fully convolutional neural network based structured
prediction approach towards the retinal vessel segmentation 2017 IEEE 14th Int. Symp. on
Biomedical Imaging (ISBI 2017) (Piscataway, NJ: IEEE) pp 248–51
[288] Fu H, Xu Y, Wong D W K and Liu J 2016 Retinal vessel segmentation via deep learning
network and fully-connected conditional random fields 2016 IEEE 13th Int. Symp. on
Biomedical Imaging (ISBI) (Piscataway, NJ: IEEE) pp 698–701
[289] Mo J and Zhang L 2017 Multi-level deep supervised networks for retinal vessel segmenta-
tion Int. J. Comput. Ass. Rad. Surg. 12 2181–93
[290] Lahiri A, Roy A G, Sheet D and Biswas P K 2016 Deep neural ensemble for retinal vessel
segmentation in fundus images towards achieving label-free angiography 2016 38th Annual

12-47
 Photo Acoustic and Optical Coherence Tomography Imaging, Volume 3

Int. Conf. of the IEEE Engineering in Medicine and Biology Society (EMBC) (Piscataway,
NJ: IEEE) pp 1340–3
[291] Li Q, Feng B, Xie L, Liang P, Zhang H and Wang T 2015 A cross-modality learning
approach for vessel segmentation in retinal images IEEE Trans. Med. Imaging 35 109–18
[292] Lim G, Cheng Y, Hsu W and Lee M L 2015 Integrated optic disc and cup segmentation
with deep learning 2015 IEEE 27th Int. Conf. on Tools with Artificial Intelligence (ICTAI)
(Piscataway, NJ: IEEE) pp 162–9
[293] Zhang D, Zhu W, Zhao H, Shi F and Chen X 2018 Automatic localization and
segmentation of optical disk based on faster r-cnn and level set in fundus image Medical
Imaging 2018: Image Processing vol 10574 (Bellingham, WA: International Society for
Optics and Photonics) p 105741U
[294] Sevastopolsky A 2017 Optic disc and cup segmentation methods for glaucoma detection
with modification of u-net convolutional neural network Pattern Recognit. Image Anal. 27
618–24
[295] Alghamdi H S, Tang H L, Waheeb S A and Peto T 2016 Automatic optic disc abnormality
detection in fundus images: a deep learning approach Proc. OMIA3 (MICCAI) pp 17–24
[296] Haloi M 2015 Improved microaneurysm detection using deep neural networks. arXiv
preprint arXiv:150504424
[297] Chudzik P, Majumdar S, Calivá F, Al-Diri B and Hunter A 2018 Microaneurysm detection
using fully convolutional neural networks Comput. Methods Programs Biomed. 158 185–92
[298] Shan J and Li L 2016 A deep learning method for microaneurysm detection in fundus
images 2016 IEEE First Int. Conf. on Connected Health: Applications, Systems and
Engineering Technologies (CHASE) (Piscataway, NJ: IEEE) pp 357–8
[299] Lahmiri S 2020 Hybrid deep learning convolutional neural networks and optimal nonlinear
support vector machine to detect presence of hemorrhage in retina Biomed. Signal Process.
Control 60 101978
[300] Derwin D J, Selvi S T and Singh O J 2020 Secondary observer system for detection of
microaneurysms in fundus images using texture descriptors J. Digit. Imaging 33 159–67
[301] Gondal W M, Köhler J M, Grzeszick R, Fink G A and Hirsch M 2017 Weakly-supervised
localization of diabetic retinopathy lesions in retinal fundus images 2017 IEEE Int. Conf.
on Image Processing (ICIP) (Piscataway, NJ: IEEE) pp 2069–73
[302] Benzamin A and Chakraborty C 2018 Detection of hard exudates in retinal fundus images
using deep learning 2018 Joint 7th Int. Conf. on Informatics, Electronics & Vision (ICIEV)
and 2018 2nd Int. Conf. on Imaging, Vision & Pattern Recognition (icIVPR). (Piscataway,
NJ: IEEE) pp 465–9
[303] Prentasic P and Loncaric S 2016 Detection of exudates in fundus photographs using deep
neural networks and anatomical landmark detection fusion Comput. Methods Programs
Biomed. 137 281–92
[304] Asiri N, Hussain M, Al Adel F and Alzaidi N 2019 Deep learning based computer-aided
diagnosis systems for diabetic retinopathy: a survey Artif. Intell. Med. 99 101701
[305] Simonyan K and Zisserman A 2014 Very deep convolutional networks for large-scale image
recognition. arXiv preprint arXiv:14091556

12-48