Practical Image Analysis

for Biology
Ofra Golani, Reinat Nevo, Slava Kalchenko
Yoseph Addadi

May 9th-24th, 2018

Weizmann Institute of Science, Rehovot, Israel
Course Aims
• Better understand digital microscopy
images, and become comfortable working
with them

• Introduce fundamental techniques in

image processing & analysis

• Practice combining these techniques to

solve problems encountered at daily work
Detailed Course Aims
• Understand the proper image acquisition needed for image
analysis.
• Understand basic concepts of image processing and analysis
• Display various types of (multi-dimensional) image datasets
• Perform manual and automatic measurements
• Automate workflows in Fiji software using scripts.
• Perform object segmentation and quantification and track
objects
• Perform pixel based and object based colocalization analysis.
• Understand why Deconvolution is needed and how to apply it
to your data.
• Get familiar with: ImageJ/Fiji, Imaris
• Understand Image Ethics
 Schedule Wednesday, 9.5
Course Intro + Basic
Thursday 10.5
L+P: Filters, Noise
Terms in Image Analysis removal, Intensity
(OG) Measurements (VK)

L+P: Digital Image L+P: Segmentation Part I

Basics, Color, (RN)
Geometrical
Measurements (VK)

Sunday, 13.5 Monday 14.5 Tuesday 15.5 Wednesday, 16.5 Thursday 17.5
L+P: Segmentation L+P: Filament Analysis L+P: 3D Display & L+P: Colocalization L+P: Fiji Macro I (OG)
part II (RN) (OG) Segmentation (OG) Analysis (RN)
BYOD: Data Presentation
L+P: Multi- L: Image Ethics (VK) L: Deconvolution (OG) L+P: Tracking (OG)
Dimensional data
(OG) L: How to Acquire
images correctly for
Image analysis (YA)

Tuesday 22.5 Wednesday, 23.5 Thursday 24.5

L+P: Fiji Macro I (OG) L+P: Fiji Macro I (OG) BYOD - work +
presentation
BYOD: Data Presentation BYOD: Data
Presentation Course – Summary

L=Lecture, P=Practice = Practice, Otherwise: Just Listen

Requirements
1 Credit point

3 Homework Assignments:
- Fiji Basics HW
- Imaris HW and/or Fiji Macro HW
- Submission until June 7th via email to
[email protected]
Use title: “HWn FirstName LastName”

Completion of the Assignments and Attendance of ALL

lectures and practice sessions is Mandatory
Classroom will be reserved for homework upon request – let us know
Administration
Communication via emails

Hands-on material, homework and lecture printouts

will be distributed through a dedicated Box folder

Recommended to use the same computer

throughout the course

At the end of this lecture we will copy the data

needed for the first days
Bring-Your-Own-Data
• Aim
– Give you a head-start with your data
– Don’t expect to solve your problem within this session
– See other problems, learn how to approach them
• How
– Short Presentations + Discussion (17.5, 22.5, 23.5)
• 5-10 minutes each
• Be focused !
– Work-on-your-own: 2-3 hours, 24.5
• We are around for questions
• Results presentations
– Better to team up in pairs
– What if I don’t have data?
Bring-Your-Own-Data
Clearly present your data and problem

– VERY short biological background – this is not the focus

– Describe the data:
• Size: 2D / 2D+t / 3D / 3D+t
• Channels: which structure is shown by each channel
– Describe your goals:
• Visualization or quantification ?
• What do you want to quantify ?
• You’ll see many examples through out the course, try to use the
terminology we use, it will make it easier for others to understand
• Do you want to do per image or per object (cell) analysis ?
• 2D or 3D analysis ?
– Time is limited, Be Focused !
– Please share with us the data.
Questions

?
 Introduction to
Quantitative Biological Image
Analysis
Ofra Golani

Images by: Li-Av Segev Zarko, Yechiel Shai , F. Cappiello et al, Antimicrob. Agents Chemother. Dec 2016 vol. 60
Imaging modalities

Confocal florescence microscopy. Source: Nature of Nature

Histology. Source: central microscopy Iowa
MRI
FIB-SEMMichal Neeman lab
Ilana Sabanay,Light microscopy:
Elior Peles, Allon Weiner Phase, DIC. Source: Nikon Microscopy
Why Quantitative Image Analysis ?

• Image data in Biology become larger and more

important
• Replace or accompany subjective visual
inspection and manual measurement by
objective quantitative computerized image
processing and analysis
• High throughput
• 10s of parameters per cell
• Reviewers require quantitative measurements
Common Image Quantifications
• What is the mean intensity of the green channel in wild-
type and knockout ?

• What is the length or area of a structure in each cell?

• How does the intensity of the GFP change over the time
course ?

• How is the protein distributed, perhaps relative to a

particular structure?

• How many cells, spots … are in the image?

• How fast and far does the object move ?

• How much does the green colocalize with the red?

Segmentation
Colocalization
Tracking
Tracing Registration

Deconvolution
Morphology
Stitiching
 Edge Detection
Filtering
Segmentation
Watershed Colocalization
Tracking
Tracing Registration
Skeleton
Deconvolution
Morphology
Stitiching
Background Subtraction
 Edge Detection
Filtering
Segmentation
HistoQuant
Watershed Colocalization
Tracking
Tracing Registration
Skeleton
Deconvolution
Morphology
Stitiching Analyze

Background Subtraction
Typical Image Analysis Workflow
• Break the analysis into
Preprocessing: Image enhancement sequential steps

• In each step: choose from

Identify Objects: Segmentation, Tracking modules of the
corresponding category

Quantification: • Interactive optimization of the

Morphology, Fluorescence, Movement analysis steps and tune user-
defined parameters

Inspection and Manual Editing, • Automation: create pipe-line

Filter, Mask, Further Analysis looping on many images

• Statistics: Group results and

Compare

• Visualization and Inspection

(sometimes correction)
BIA Components and Workflows

Sebastien Tosi, Romain Guiet, NEUBIAS-WG5

Segmentation
Colocalization
Tracking
Tracing Registration

Deconvolution
Morphology
Stitiching
Segmentation

Epithelial cells, Florescence microscopy. Tal Ilani and Deborah Fass

Segmentation

Detect Cells

Detection: counting only

Epithelial cells, Florescence microscopy. Tal Ilani and Deborah Fass

Segmentation

Segment Cells

Measure Cells

Segmentation: Counting, Measurement, Filtering

Epithelial cells, Florescence microscopy. Tal Ilani and Deborah Fass

 Measurements
Intensity Morphology

For each channel: Mean, Std

Ratio between channels

For each object

Distribution

Naffar-Abu-Amara S, Shay T, Galun M, Cohen N, Isakoff SJ, Kam Z, Geiger B. Identification of Novel Pro-Migratory, Cancer-
Associated Genes Using Quantitative, Microscopy-Based Screening. PloS One, 3 2008
Relations between Objects
Relation between objects and parts of objects
• Cells and Nuclei or vesicles, Tissue and cells
• Number/density of vesicles,
• Measure Intensity in nuclei/cytoplasm

Relation between objects of the same type

• Distances
Cell with 2 population of vesicles
• Spatial order Bitplane website

• Clustering

Lagache T et al. PLoS ONE 8(12)

Relation between objects of different populations

• Distances
• Colocalization Quantifying bacteria internalization by macrophage
Katia Fettucciari et al. Cellular Microbiology, Vol. 13,
Issue 6
Time-lapse - Tracking objects
Target: Inspect live objects over time

• Follow the same object over time,

and measure changes in this object

• Measure Intensity,
speed, change of speed
over time,
direction of movement

• Detect cell division,

track lineages
Tracking: Dynamic Proteomics

A A Cohen et al. Science 2008;322:1511-1516

Tracking: Dynamic Proteomics

Why is it hard ?

• Contrast
• Non rigid cell shapes
• Cells cross each other
• Sporadic entry and exit out of the field of
view
• Cluttering and occlusion
• Cells undergo dramatic physiological
changes (e.g. cell explosions and death)
 Tracking: Dynamic Proteomics
Design Biological setup that will
support automatic quantitative
image analysis
Couple it with sophisticated Image
analysis

Key Issues:
Segmentation is done using different
channel than measurements of
Intensity (gene expression)

Red florescent channel is used for

segmentation - nuclei and cytoplasm
at different intensities
Yellow florescent is used for protein
level and localization
A A Cohen et al. Science 2008;322:1511-1516
Tracking: Dynamic Proteomics

A A Cohen et al. Science 2008;322:1511-1516

Tracking: Dynamic Proteomics

Rapid translocations in
response to the drug
correspond to
nucleolar stress and
oxidative stress
pathways.

A A Cohen et al. Science 2008;322:1511-1516

Filament / Fiber Analysis
Quantification of fibrous objects (thin lines)
Neurons, microtubules, blood vessels
 Advanced Preprocessing: Deconvolution

Before (left) and After (right) AutoQuant. Image courtesy of Richard Cole, NYS
Department of Health, Biggs Laboratory- Wadsworth Center, Albany NY, USA.
Advanced Preprocessing: Image Registration

• Transforming different sets

of data into one coordinate
system

• Data may be multiple

images from different
sensors, times, depths, or
viewpoints

• Necessary to compare or
integrate the data obtained Measurements from different
from different Modalities: 3D registration of PET
functional data to MRI structural
measurements data.
Source: Analyze Direct
Image Registration
Independent movement of different organs -
Registration of Whole Volume

Reut Avni and Michal Neeman

Image Registration
Independent movement of different organs -
Registration of Whole Volume

Reut Avni and Michal Neeman

Image Registration
Independent movement of different organs -
Registration of Volume-Of-Interest

Reut Avni and Michal Neeman

Image Registration
Independent movement of different organs -
Registration of Volume-Of-Interest

Reut Avni and Michal Neeman

Image Analysis Tools
• Hard to Standardize
• Adjust and tailor for every problem
• Tune parameters and build pipelines
• Commercial / Free tools
• Interface for Extensions
• Develop new tools for specific applications

Analyze

HistoQuant

…
Useful Resources
“Analyzing fluorescence microscopy images
with ImageJ” by Peter Bankhead.
Original PDF or Updated GitBook

BIAS E-Book:
– Based on "The EMBL Master Course for Bioimage Data
Analysis"
– Chapter 3: ImageJ Macro Language
– Learn through Workflows
- Bioimage Data Analysis (2016), Wiley-VCH
- Will be updated continuously online !
Find your way
• Join ImageJ Forum & ImageJ mailing list
• Biii.eu: A good starting point
• Contact us for help
Get Prepared
Open your email, Get link to Box folder

Download the Hands-on data from the Box folder to

D:\ImageAnalysisCourse-2018