Download Complete Computational Analysis and Deep Learning for Medical Care Principles Methods and Applications 1st Edition Amit Kumar Tyagi Editor PDF for All Chapters
Download Complete Computational Analysis and Deep Learning for Medical Care Principles Methods and Applications 1st Edition Amit Kumar Tyagi Editor PDF for All Chapters
com
https://ebookmeta.com/product/computational-analysis-and-
deep-learning-for-medical-care-principles-methods-and-
applications-1st-edition-amit-kumar-tyagi-editor-2/
OR CLICK HERE
DOWLOAD NOW
https://ebookmeta.com/product/data-science-for-genomics-1st-edition-
amit-kumar-tyagi/
ebookmeta.com
https://ebookmeta.com/product/computational-methods-for-deep-learning-
theoretic-practice-and-applications-texts-in-computer-science-wei-qi-
yan/
ebookmeta.com
https://ebookmeta.com/product/diatom-gliding-motility-biology-and-
applications-1st-edition-kalina-m-manoylov/
ebookmeta.com
Demon Lord Retry Volume 2 1st Edition Kurone Kanzaki
https://ebookmeta.com/product/demon-lord-retry-volume-2-1st-edition-
kurone-kanzaki/
ebookmeta.com
https://ebookmeta.com/product/endurance-sports-medicine-a-clinical-
guide-timothy-l-miller/
ebookmeta.com
https://ebookmeta.com/product/fundamentals-of-structural-mechanics-
dynamics-and-stability-1st-edition-a-i-rusakov/
ebookmeta.com
https://ebookmeta.com/product/envisioning-black-feminist-voodoo-
aesthetics-african-spirituality-in-american-cinema-kameelah-l-martin/
ebookmeta.com
https://ebookmeta.com/product/introduction-to-bronchoscopy-2nd-
edition-armin-ernst-felixj-f-herth/
ebookmeta.com
Table of Contents
Cover
Title Page
Copyright
Preface
Part 1: Deep Learning and Its Models
1 CNN: A Review of Models, Application of IVD Segmentation
1.1 Introduction
1.2 Various CNN Models
1.3 Application of CNN to IVD Detection
1.4 Comparison With State-of-the-Art Segmentation
Approaches for Spine T2W Images
1.5 Conclusion
References
2 Location-Aware Keyword Query Suggestion Techniques
With Artificial Intelligence Perspective
2.1 Introduction
2.2 Related Work
2.3 Artificial Intelligence Perspective
2.4 Architecture
2.5 Conclusion
References
3 Identification of a Suitable Transfer Learning Architecture
for Classification: A Case Study with Liver Tumors
3.1 Introduction
3.2 Related Works
3.3 Convolutional Neural Networks
3.4 Transfer Learning
3.5 System Model
3.6 Results and Discussions
3.7 Conclusion
References
4 Optimization and Deep Learning-Based Content Retrieval,
Indexing, and Metric Learning Approach for Medical Images
4.1 Introduction
4.2 Related Works
4.3 Proposed Method
4.4 Results and Discussion
4.5 Conclusion
References
Part 2: Applications of Deep Learning
5 Deep Learning for Clinical and Health Informatics
5.1 Introduction
5.2 Related Work
5.3 Motivation
5.4 Scope of the Work in Past, Present, and Future
5.5 Deep Learning Tools, Methods Available for Clinical,
and Health Informatics
5.6 Deep Learning: Not-So-Near Future in Biomedical
Imaging
5.7 Challenges Faced Toward Deep Learning Using in
Biomedical Imaging
5.8 Open Research Issues and Future Research Directions
in Biomedical Imaging (Healthcare Informatics)
5.9 Conclusion
References
6 Biomedical Image Segmentation by Deep Learning Methods
6.1 Introduction
6.2 Overview of Deep Learning Algorithms
6.3 Other Deep Learning Architecture
6.4 Biomedical Image Segmentation
6.5 Conclusion
References
7 Multi-Lingual Handwritten Character Recognition Using
Deep Learning
7.1 Introduction
7.2 Related Works
7.3 Materials and Methods
7.4 Experiments and Results
7.5 Conclusion
References
8 Disease Detection Platform Using Image Processing Through
OpenCV
8.1 Introduction
8.2 Problem Statement
8.3 Conclusion
8.4 Summary
References
9 Computer-Aided Diagnosis of Liver Fibrosis in Hepatitis
Patients Using Convolutional Neural Network
9.1 Introduction
9.2 Overview of System
9.3 Methodology
9.4 Performance and Analysis
9.5 Experimental Results
9.6 Conclusion and Future Scope
References
Part 3: Future Deep Learning Models
10 Lung Cancer Prediction in Deep Learning Perspective
10.1 Introduction
10.2 Machine Learning and Its Application
10.3 Related Work
10.4 Why Deep Learning on Top of Machine Learning?
10.5 How is Deep Learning Used for Prediction of Lungs
Cancer?
10.6 Conclusion
References
11 Lesion Detection and Classification for Breast Cancer
Diagnosis Based on Deep CNNs from Digital Mammographic
Data
11.1 Introduction
11.2 Background
11.3 Methods
11.4 Application of Deep CNN for Mammography
11.5 System Model and Results
11.6 Research Challenges and Discussion on Future
Directions
11.7 Conclusion
References
12 Health Prediction Analytics Using Deep Learning Methods
and Applications
12.1 Introduction
12.2 Background
12.3 Predictive Analytics
12.4 Deep Learning Predictive Analysis Applications
12.5 Discussion
12.6 Conclusion
References
13 Ambient-Assisted Living of Disabled Elderly in an
Intelligent Home Using Behavior Prediction—A Reliable Deep
Learning Prediction System
13.1 Introduction
13.2 Activities of Daily Living and Behavior Analysis
13.3 Intelligent Home Architecture
13.4 Methodology
13.5 Senior Analytics Care Model
13.6 Results and Discussions
13.7 Conclusion
Nomenclature
References
14 Early Diagnosis Tool for Alzheimer’s Disease Using 3D
Slicer
14.1 Introduction
14.2 Related Work
14.3 Existing System
14.4 Proposed System
14.5 Results and Discussion
14.6 Conclusion
References
Part 4: Deep Learning - Importance and Challenges for Other
Sectors
15 Deep Learning for Medical Healthcare: Issues, Challenges,
and Opportunities
15.1 Introduction
15.2 Related Work
15.3 Development of Personalized Medicine Using Deep
Learning: A New Revolution in Healthcare Industry
15.4 Deep Learning Applications in Precision Medicine
15.5 Deep Learning for Medical Imaging
15.6 Drug Discovery and Development: A Promise
Fulfilled by Deep Learning Technology
15.7 Application Areas of Deep Learning in Healthcare
15.8 Privacy Issues Arising With the Usage of Deep
Learning in Healthcare
15.9 Challenges and Opportunities in Healthcare Using
Deep Learning
15.10 Conclusion and Future Scope
References
16 A Perspective Analysis of Regularization and Optimization
Techniques in Machine Learning
16.1 Introduction
16.2 Regularization in Machine Learning
16.3 Convexity Principles
16.4 Conclusion and Discussion
References
17 Deep Learning-Based Prediction Techniques for Medical
Care: Opportunities and Challenges
17.1 Introduction
17.2 Machine Learning and Deep Learning Framework
17.3 Challenges and Opportunities
17.4 Clinical Databases—Electronic Health Records
17.5 Data Analytics Models—Classifiers and Clusters
17.6 Deep Learning Approaches and Association
Predictions
17.7 Conclusion
17.8 Applications
References
18 Machine Learning and Deep Learning: Open Issues and
Future Research Directions for the Next 10 Years
18.1 Introduction
18.2 Evolution of Machine Learning and Deep Learning
18.3 The Forefront of Machine Learning Technology
18.4 The Challenges Facing Machine Learning and Deep
Learning
18.5 Possibilities With Machine Learning and Deep
Learning
18.6 Potential Limitations of Machine Learning and Deep
Learning
18.7 Conclusion
Acknowledgement
Contribution/Disclosure
References
Index
List of Illustrations
Chapter 1
Figure 1.1 Architecture of LeNet-5.
Figure 1.2 Architecture of AlexNet.
Figure 1.3 Architecture of ZFNet.
Figure 1.4 Architecture of VGG-16.
Figure 1.5 Inception module.
Figure 1.6 Architecture of GoogleNet.
Figure 1.7 (a) A residual block.
Figure 1.8 Architecture of ResNeXt.
Figure 1.9 Architecture of SE-ResNet.
Figure 1.10 Architecture of DenseNet.
Figure 1.11 Architecture of MobileNets.
Chapter 2
Figure 2.1 General architecture of a search engine.
Figure 2.2 The increased mobile users.
Figure 2.3 AI-powered location-based system.
Figure 2.4 Architecture diagram for querying.
Chapter 3
Figure 3.1 Phases of CECT images (1: normal liver; 2: tumor
within liver; 3: sto...
Figure 3.2 Architecture of convolutional neural network.
Figure 3.3 AlexNet architecture.
Figure 3.4 GoogLeNet architecture.
Figure 3.5 Residual learning—building block.
Figure 3.6 Architecture of ResNet-18.
Figure 3.7 System model for case study on liver tumor
diagnosis.
Figure 3.8 Output of bidirectional region growing
segmentation algorithm: (a) in...
Figure 3.9 HA Phase Liver CT images: (a) normal liver; (b)
HCC; (c) hemangioma; ...
Figure 3.10 Training progress for AlexNet.
Figure 3.11 Training progress for GoogLeNet.
Figure 3.12 Training progress for ResNet-18.
Figure 3.13 Training progress for ResNet-50.
Chapter 4
Figure 4.1 Proposed system for image retrieval.
Figure 4.2 Schematic of the deep convolutional neural
networks.
Figure 4.3 Proposed feature extraction system.
Figure 4.4 Proposed model for the localization of the
abnormalities.
Figure 4.5 Graph for the retrieval performance of the metric
learning for VGG19.
Figure 4.6 PR values for state of art ConvNet model for CT
images.
Figure 4.7 PR values for state of art CNN model for CT images.
Figure 4.8 Proposed system—PR values for the CT images.
Figure 4.9 PR values for proposed content-based image
retrieval.
Figure 4.10 Graph for loss function of proposed deep
regression networks for tra...
Figure 4.11 Graph for loss function of proposed deep
regression networks for val...
Chapter 6
Figure 5.1 Different informatics in healthcare [28].
Chapter 6
Figure 6.1 CT image reconstruction (past, present, and future)
[3].
Figure 6.2 (a) Classic machine learning algorithm, (b) Deep
learning algorithm.
Figure 6.3 Traditional neural network.
Figure 6.4 Convolutional Neural Network.
Figure 6.5 Psoriasis images [2].
Figure 6.6 Restricted Boltzmann Machine.
Figure 6.7 Autoencoder architecture with vector and image
inputs [1].
Figure 6.8 Image of chest x-ray [60].
Figure 6.9 Regular thoracic disease identified in chest x-rays
[23].
Figure 6.10 MRI of human brain [4].
Chapter 7
Figure 7.1 Architecture of the proposed approach.
Figure 7.2 Sample Math dataset (including English
characters).
Figure 7.3 Sample Bangla dataset (including Bangla numeric).
Figure 7.4 Sample Devanagari dataset (including Hindi
numeric).
Figure 7.5 Dataset distribution for English dataset.
Figure 7.6 Dataset distribution for Hindi dataset.
Figure 7.7 Dataset distribution for Bangla dataset.
Figure 7.8 Dataset distribution for Math Symbol dataset.
Figure 7.9 Dataset distribution.
Figure 7.10 Precision-recall curve on English dataset.
Figure 7.11 ROC curve on English dataset.
Figure 7.12 Precision-recall curve on Hindi dataset.
Figure 7.13 ROC curve on Hindi dataset.
Figure 7.14 Precision-recall curve on Bangla dataset.
Figure 7.15 ROC curve on Bangla dataset.
Figure 7.16 Precision-recall curve on Math Symbol dataset.
Figure 7.17 ROC curve on Math symbol dataset.
Figure 7.18 Precision-recall curve of the proposed model.
Figure 7.19 ROC curve of the proposed model.
Chapter 8
Figure 8.1 Eye image dissection [34].
Figure 8.2 Cataract algorithm [10].
Figure 8.3 Pre-processing algorithm [48].
Figure 8.4 Pre-processing analysis [39].
Figure 8.5 Morphologically opened [39].
Figure 8.6 Finding circles [40].
Figure 8.7 Iris contour separation [40].
Figure 8.8 Image inversion [41].
Figure 8.9 Iris detection [41].
Figure 8.10 Cataract detection [41].
Figure 8.11 Healthy eye vs. retinoblastoma [33].
Figure 8.12 Unilateral retinoblastoma [18].
Figure 8.13 Bilateral retinoblastoma [19].
Figure 8.14 Classification of stages of skin cancer [20].
Figure 8.15 Eye cancer detection algorithm.
Figure 8.16 Sample test cases.
Figure 8.17 Actual working of the eye cancer detection
algorithm.
Figure 8.18 Melanoma example [27].
Figure 8.19 Melanoma detection algorithm.
Figure 8.20 Asymmetry analysis.
Figure 8.21 Border analysis.
Figure 8.22 Color analysis.
Figure 8.23 Diameter analysis.
Figure 8.24 Completed detailed algorithm.
Chapter 9
Figure 9.1 Basic overview of a proposed computer-aided
system.
Figure 9.2 Block diagram of the proposed system for finding
out liver fibrosis.
Figure 9.3 Block diagram representing different pre-
processing stages in liver f...
Figure 9.4 Flow chart showing student’s t test.
Figure 9.5 Diagram showing SegNet architecture for
convolutional encoder and dec...
Figure 9.6 Basic block diagram of VGG-16 architecture.
Figure 9.7 Flow chart showing SegNet working process for
classifying liver fibro...
Figure 9.8 Overall process of the CNN of the system.
Figure 9.9 The stages in identifying liver fibrosis by using
Conventional Neural...
Figure 9.10 Multi-layer neural network architecture for a CAD
system for diagnos...
Figure 9.11 Graphical representation of Support Vector
Machine.
Figure 9.12 Experimental analysis graph for different
classifier in terms of acc...
Chapter 10
Figure 10.1 Block diagram of machine learning.
Figure 10.2 Machine learning algorithm.
Figure 10.3 Structure of deep learning.
Figure 10.4 Architecture of DNN.
Figure 10.5 Architecture of CNN.
Figure 10.6 System architecture.
Figure 10.7 Image before histogram equalization.
Figure 10.8 Image after histogram equalization.
Figure 10.9 Edge detection.
Figure 10.10 Edge segmented image.
Figure 10.11 Total cases.
Figure 10.12 Result comparison.
Chapter 11
Figure 11.1 Breast cancer incidence rates worldwide (source:
International Agenc...
Figure 11.2 Images from MIAS database showing normal,
benign, malignant mammogra...
Figure 11.3 Image depicting noise in a mammogram.
Figure 11.4 Architecture of CNN.
Figure 11.5 A complete representation of all the operation
that take place at va...
Figure 11.6 An image depicting Pouter, Plesion, and Pbreast in
a mammogram.
Figure 11.7 The figure depicts two images: (a) mammogram
with a malignant mass a...
Figure 11.8 A figure depicting the various components of a
breast as identified ...
Figure 11.9 An illustration of how a mammogram image
having tumor is segmented t...
Figure 11.10 A schematic representation of classification
procedure of CNN.
Figure 11.11 A schematic representation of classification
procedure of CNN durin...
Figure 11.12 Proposed system model.
Figure 11.13 Flowchart for MIAS database and unannotated
labeled images.
Figure 11.14 Image distribution for training model.
Figure 11.15 The graph shows the loss for the trained model
on train and test da...
Figure 11.16 The graph shows the accuracy of the trained
model for both test and...
Figure 11.17 Depiction of the confusion matrix for the trained
CNN model.
Figure 11.18 Receiver operating characteristics of the trained
model.
Figure 11.19 The image shows the summary of the CNN
model.
Figure 11.20 Performance parameters of the trained model.
Figure 11.21 Prediction of one of the image collected from
diagnostic center.
Chapter 12
Figure 12.1 Deep learning [14]. (a) A simple, multilayer deep
neural network tha...
Figure 12.2 Flowchart of the model [25]. The orange icon
indicates the dataset, ...
Figure 12.3 Evaluation result [25].
Figure 12.4 Deep learning techniques evaluation results [25].
Figure 12.5 Deep transfer learning–based screening system
[38].
Figure 12.6 Classification result.
Figure 12.7 Regression result [45].
Figure 12.8 AE model of deep learning [47].
Figure 12.9 DBN for induction motor fault diagnosis [68].
Figure 12.10 CNN model for health monitoring [80].
Figure 12.11 RNN model for health monitoring [87].
Figure 12.12 Deep learning models usage.
Chapter 13
Figure 13.1 Intelligent home layout model.
Figure 13.2 Deep learning model in predicting behavior
analysis.
Figure 13.3 Lifestyle-oriented context aware model.
Figure 13.4 Components for the identification, simulation, and
detection of acti...
Figure 13.5 Prediction stages.
Figure 13.6 Analytics of event.
Figure 13.7 Prediction of activity duration.
Chapter 14
Figure 14.1 Comparison of normal and Alzheimer brain.
Figure 14.2 Proposed AD prediction system.
Figure 14.3 KNN classification.
Figure 14.4 SVM classification.
Figure 14.5 Load data in 3D slicer.
Figure 14.6 3D slicer visualization.
Figure 14.7 Normal patient MRI.
Figure 14.8 Alzheimer patient MRI.
Figure 14.9 Comparison of hippocampus region.
Figure 14.10 Accuracy of algorithms with baseline records.
Figure 14.11 Accuracy of algorithms with current records.
Figure 14.12 Comparison of without and with dice coefficient.
Chapter 15
Figure 15.1 U-Net architecture [19].
Figure 15.2 Architecture of the 3D-DCSRN model [29].
Figure 15.3 SMILES code for Cyclohexane and Acetaminophen
[32].
Figure 15.4 Medical chatbot architecture [36].
Chapter 16
Figure 16.1 A classical perceptron.
Figure 16.2 Forward and backward paths on an ANN
architecture.
Figure 16.3 A DNN architecture.
Figure 16.4 A DNN architecture for digit classification.
Figure 16.5 Underfit and overfit.
Figure 16.6 Functional mapping.
Figure 16.7 A generalized Tikhonov functional.
Figure 16.8 (a) With hidden layers (b) Dropping h2 and h5.
Figure 16.9 Image cropping as one of the features of data
augmentation.
Figure 16.10 Early stopping criteria based on errors.
Figure 16.11 (a) Convex, (b) Non-convex.
Figure 16.12 (a) Affine (b) Convex function.
Figure 16.13 Workflow and an optimizer.
Figure 16.14 (a) Error (cost) function (b) Elliptical: Horizontal
cross section.
Figure 16.15 Contour plot for a quadratic cost function with
elliptical contours...
Figure 16.16 Gradients when steps are varying.
Figure 16.17 Local minima. (When the gradient ∇ of the
partial derivatives is po...
Figure 16.18 Contour plot showing basins of attraction.
Figure 16.19 (a) Saddle point S. (b) Saddle point over a two-
dimensional error s...
Figure 16.20 Local information encoded by the gradient
usually does not support ...
Figure 16.21 Direction of gradient change.
Figure 16.22 Rolling ball and its trajectory.
Chapter 17
Figure 17.1 Artificial Neural Networks vs. Architecture of
Deep Learning Model [...
Figure 17.2 Machine learning and deep learning techniques
[4, 5].
Figure 17.3 Model of reinforcement learning
(https://www.kdnuggets.com).
Figure 17.4 Data analytical model [5].
Figure 17.5 Support Vector Machine—classification approach
[1].
Figure 17.6 Expected output of K-means clustering [1].
Figure 17.7 Output of mean shift clustering [2].
Figure 17.8 Genetic Signature–based Hierarchical Random
Forest Cluster (G-HR Clu...
Figure 17.9 Artificial Neural Networks vs. Deep Learning
Neural Networks.
Figure 17.10 Architecture of Convolution Neural Network.
Figure 17.11 Architecture of the Human Diseases Pattern
Prediction Technique (EC...
Figure 17.12 Comparative analysis: processing time vs.
classifiers.
Figure 17.13 Comparative analysis: memory usage vs.
classifiers.
Figure 17.14 Comparative analysis: classification accuracy vs.
classifiers.
Figure 17.15 Comparative analysis: sensitivity vs. classifiers.
Figure 17.16 Comparative analysis: specificity vs. classifiers.
Figure 17.17 Comparative analysis: FScore vs. classifiers.
Chapter 18
Figure 18.1 Deep Neural Network (DNN).
Figure 18.2 The evolution of machine learning techniques
(year-wise).
List of Tables
Chapter 1
Table 1.1 Various parameters of the layers of LeNet.
Table 1.2 Every column indicates which feature map in S2 are
combined by the uni...
Table 1.3 AlexNet layer details.
Table 1.4 Various parameters of ZFNet.
Table 1.5 Various parameters of VGG-16.
Table 1.6 Various parameters of GoogleNet.
Table 1.7 Various parameters of ResNet.
Table 1.8 Comparison of ResNet-50 and ResNext-50 (32 × 4d).
Table 1.9 Comparison of ResNet-50 and ResNext-50 and SE-
ResNeXt-50 (32 × 4d).
Table 1.10 Comparison of DenseNet.
Table 1.11 Various parameters of MobileNets.
Table 1.12 State-of-art of spine segmentation approaches.
Chapter 2
Table 2.1 History of search engines.
Table 2.2 Three types of user refinement of queries.
Table 2.3 Different approaches for the query suggestion
techniques.
Chapter 3
Table 3.1 Types of liver lesions.
Table 3.2 Dataset count.
Table 3.3 Hyperparameter settings for training.
Table 3.4 Confusion matrix for AlexNet.
Table 3.5 Confusion matrix for GoogLeNet.
Table 3.6 Confusion matrix for ResNet-18.
Table 3.7 Confusion matrix for ResNet-50.
Table 3.8 Comparison of classification accuracies.
Chapter 4
Table 4.1 Retrieval performance of metric learning for VGG19.
Table 4.2 Performance of retrieval techniques of the trained
VGG19 among fine-tu...
Table 4.3 PR values of various models—a comparison for CT
image retrieval.
Table 4.4 Recall vs. precision for proposed content-based
image retrieval.
Table 4.5 Loss function of proposed deep regression networks
for training datase...
Table 4.6 Loss function of proposed deep regression networks
for validation data...
Table 4.7 Land mark details (identification rates vs. distance
error) for the pr...
Table 4.8 Accuracy value of the proposed system.
Table 4.9 Accuracy of the retrieval methods compared with
the metric learning–ba...
Chapter 6
Table 6.1 Definition of the abbreviations.
Chapter 7
Table 7.1 Performance of proposed models on English dataset.
Table 7.2 Performance of proposed model on Bangla dataset.
Table 7.3 Performance of proposed model on Math Symbol
dataset.
Chapter 8
Table 8.1 ABCD factor for TDS value.
Table 8.2 Classify mole according to TDS value.
Chapter 9
Table 9.1 The confusion matrix for different classifier.
Table 9.2 Performance analysis of different classifiers:
Random Forest, SVM, Naï...
Chapter 10
Table 10.1 Result analysis.
Chapter 11
Table 11.1 Comparison of different techniques and tumor.
Chapter 13
Table 13.1 Cognitive functions related with routine activities.
Table 13.2 Situation and design features.
Table 13.3 Accuracy of prediction.
Chapter 14
Table 14.1 Accuracy comparison and mean of algorithms with
baseline records.
Table 14.2 Accuracy comparison and mean of algorithms with
current records.
Chapter 15
Table 15.1 Variances of Convolutional Neural Network (CNN).
Table 15.2 Various issues challenges faced by researchers for
using deep learnin...
Chapter 17
Table 17.1 Comparative analysis: classification accuracy for 10
datasets—analysi...
Chapter 18
Table 18.1 Comparison among data mining, machine learning,
and deep learning.
Scrivener Publishing
100 Cummings Center, Suite 541J
Beverly, MA 01915-6106
Publishers at Scrivener
Martin Scrivener ([email protected])
Phillip Carmical ([email protected])
Computational Analysis and
Deep Learning for Medical
Care
Abstract
The widespread publicity of Convolutional Neural Network (CNN) in various domains such as image
classification, object recognition, and scene classification has revolutionized the research in machine
learning, especially in medical images. Magnetic Resonance Images (MRIs) are suffering from severe
noise, weak edges, low contrast, and intensity inhomogeneity. Recent advances in deep learning with
fewer connections and parameters made their training easier. This chapter presents an in-depth review
of the various deep architectures as well as its application for segmenting the Intervertebral disc (IVD)
from the 3D spine image and its evaluation. The first section deals with the study of various traditional
architectures of deep CNN such as LeNet, AlexNet, ZFNet, GoogleNet, VGGNet, ResNet, Inception model,
ResNeXt, SENet, MobileNet V1/V2, and DenseNet. It also deals with the study of the parameters and
components associated with the models in detail. The second section discusses the application of these
models to segment IVD from the spine image. Finally, theoretically performance and experimental results
of the state-of-art of the literature shows that 2.5D multi-scale FCN performs the best with the Dice
Similarity Index (DSC) of 90.64%.
Keywords: CNN, deep learning, intervertebral disc degeneration, MRI segmentation
1.1 Introduction
The concept of Convolutional Neural Network (CNN) was introduced by Fukushima. The principle in CNN is
that the visual mechanism of human is hierarchical in structure. CNN has been successfully applied in
various image domain such as image classification, object recognition, and scene classification. CNN is
defined as a series of convolution layer and pooling layer. In the convolution layer, the image is convolved
with a filter, i.e., slide over the image spatially and computing dot products. Pooling layer provides a smaller
feature set.
One major cause of low back pain is disc degeneration. Automated detection of lumbar abnormalities from
the clinical scan is a burden for radiologist. Researchers focus on the automation task of the segmentation of
large set of MRI data due to the huge size of such images. The success of the application of CNN in various
field of object detection enables the researchers to apply various models for the detection of Intervertebral
Disc (IVD) and, in turn, helps in the diagnosis of diseases.
The details of the structure of the remaining section of the paper are as follows. The next section deals with
the study of the various CNN models. Section 1.3, presents applications of CNN for the detection of the IVD.
In Section 1.4, comparison with state-of-the-art segmentation approaches for spine T2W images is carried
out, and conclusion is in Section 1.5.
11. For the information of such readers as have not access to large
works, I will take the liberty of inserting an account of the
dimensions of these celebrated and ancient monuments, from
Sir R. Wilson's history.
"Pompey's Pillar is of the Corinthian order, and eighty-eight feet
six inches in height; the shaft formed of a single block of
granite, retaining the finest polish, except where the wind on
the north-east front has chafed the surface a little; it is sixty-
four feet in height, and eight feet four inches in diameter.
"About thirty yards in the rear of the French intrenchments,
stands Cleopatra's Needle, and one of equal magnitude is lying
close by, horizontally. The form of these obelisks is of
considerable elegance, and their magnitude is enormous,
considering that each is only one piece of granite; their height
is sixty-eight feet three inches, and their base seven feet seven
inches by seven feet square; their sides are covered with
hieroglyphics, which, on the eastern front of the one that is
upright, are much effaced by the wind.
"Tradition affirms that they ornamented the gate of Cleopatra's
palace. From the quantity of marble, &c. &c. found near the
spot, probably the residence of the sovereigns of Egypt was
placed there."—History of the Expedition, 2d vol. pp. 156, 158,
159.
Dr. E. D. Clark, the traveller, who has paid great attention to
the study of the age and design of ancient monuments, thinks
that the shaft of Pompey's Pillar "is of much earlier antiquity
than either the capital or the pedestal." He gives probable
reasons to believe that the shaft was made in the time of
Alexander the Great, the founder of Alexandria, and who was
buried there, to be a sepulchral pillar to the memory of that
monarch; but that Julius Caesar had set it upon a pedestal,
and had put a capital upon it in honour of Pompey, whose
head he caused to be burnt with funeral honours, and the
ashes put into an urn, and placed on the top of the pillar: but
that the pillar had likely fallen afterwards, and had been
restored by the emperor Hadrian.—Clarke's Travels, 4th Edit.
8vo. vol. v. ch. vii. p. 361, &c.
14. This wind was still more dreadful in the interior of the country;
and at the place where the army was on its march to Cairo; as
appears by the following extract from Sir R. Wilson's History of
the Expedition to Egypt, vol. 1. p. 177.
ALGUM, 23d May.
"This day will ever be remarkable to the Egyptian army; a
sirocco wind darkened with a burning mist the atmosphere; the
thermometer was at 120 in the shade; the ground was heated
like the floor of a furnace; every thing that was metallic, such
as arms, buttons; knives, &c. became burning hot; the poultry,
exposed to the air, and several horses and camels died;
respiration was difficult, and the lungs were parched with fiery
particles. Had the heat continued forty-eight hours, the effect
would have been dreadful: but happily as night drew on, the
wind cooled, and at last changed to the north west.
"At Balbeis, the thermometer was at 130; on the western side
of the Nile 120; at Alexandria 105."
Extract from a Journal written by one of my comrades.
"We had one day's hot wind from the south; it began to blow
about 9 o'clock; and wo be to him that is far from shelter, as
neither man nor beast can survive it three days! It came from
the desert as hot as the opening of an oven door, bringing
small sand like mist along with it. All the sentinels were called
in, and the cattle crept close to the ground and groaned for
fear. The buffaloes took to the river, covering themselves, all
but the nose, in the water; and no man was able to stir out of
his tent until the evening."
15. The promise that I made of informing his relatives of the time
and circumstances of his death, I fulfilled when I came to
Ireland, for which I received a letter of thanks from his brother.
CHAPTER VI.
16. During the time of the inundation, the water in the river is very
thick but as much pure water as served us for drinking, was
procured from some private wells in the town, which I suppose
had a communication with the river, which had the effect of
filtering the water.
19. The grain was measured by an Arab into baskets, which were
carried to the heap by others, upon their shoulders. The
measurer accompanied his work with a song indicative of the
quantity he put into each basket. The owner stood upon the
quay and received a bean or pea from the carriers as they
passed by him to the heap; and this was the method by which
he kept an account of the quantity landed.
20. I saw the exterior of some of these houses on the banks of the
Nile, but never had an opportunity of seeing their interior. Sir
R. Wilson says, in vol. i. pp. 156, 157, "All language is
insufficient to give a just idea of the misery of an Egyptian
village; but those who have been in Ireland, may best suppose
the degree, when an Irish hut is described as a palace, in
comparison to an Arab's stye; for it can be called by no other
name. Each habitation is built of mud, even the roof, and
resembles in shape an oven: within is only one apartment,
generally of about ten feet square. The door does not admit of
a man's entering upright; but, as the bottom is dug out about
two feet, when in the room an erect posture is possible. A mat,
some large vessels to hold water, which is the constant
occupation of the women to fetch; a pitcher made of fine
porous clay, found best in Upper Egypt, near Cunei, and in
which the water is kept very cool; a rice pan and coffee pot,
are all the ornaments and utensils. Here, then, a whole family
eat and sleep without any consideration of decency or
cleanliness; being, in regard to the latter, worse even than the
beasts of the field, which naturally respect their own
tenements."
26. "Sir Sidney Smith informed the author (Dr. Clarke) that one
night, preferring a bed upon the sand of the desert to a night's
lodging in the village of Etko, as thinking he should be more
secure from vermin, he found himself, in the morning, entirely
covered by them. Lice and scorpions abound in all the sandy
desert near Alexandria." One of my comrades informed me,
that when some of the date trees were split at Aboukir, for
making the hospital, there were so many lice in the hearts of
them that they might have been gathered in handfuls. The
frogs also were so abundant at some of the places where the
army halted between Rosetta and Cairo, that it was not
possible to get at the water in the river without treading upon
them; and at one place the camp ground was literally covered
with black beetles, to the no small annoyance of the soldiers in
the tents, and the bed frames and mats that we got new in the
hospital in Rosetta in the end of June, were so full of bugs by
the end of September, that they were fit only to be burnt.
30. When their rations happened to be salt pork, they used to put
a piece of it under the kettle to burn with the straw.
31. With the exception of gold, which was in the hands of a few,
the coin circulating in Egypt was made of base metal, watered
over with silver; and was of little or no intrinsic value. There
were large pieces of this kind, some of them larger than a
crown, which were of different values: but a small coin, called
a para, about the breadth of a farthing, and no thicker than
the scale of a fish, was the most common; of which 120, and
in some places 160, were given for a Spanish dollar. The
money expended by the army was gold and Spanish dollars.
32. The inundation in this lake extended farther than the eye could
reach. The banks of the canal formed a road for
communicating with the interior of the country; a bridge of
boats united the banks, one of the boats being moveable, for
the purpose of allowing vessels to pass in and out of Lake
Mareotis. Before the army wholly left the country, the boats
forming the bridge were sunk in the cut, and served for a
foundation upon which the banks were rebuilt. When the
British took Alexandria, in March, 1807, a detachment was sent
to take Rosetta; but they were repulsed by those Turks who
had accompanied the army on its march to and from Cairo,
and who had acquired a considerable portion of British
discipline. The rays of the sun had by this time so far dried up
the salt water in Lake Mareotis as to render it passable; but
the British again cut the banks of the canal, and admitted the
sea into it, to protect Alexandria from being attacked by the
Turks.
CHAPTER VII.