An automated image analysis tool for quantification of fat cells

This tool was designed for the experiment mentioned in: [https://www.sciencedirect.com/science/article/pii/S2667160323000133]

GUI version avaliable in: [https://github.com/nehalkalita/Fatquant-GUI]

• Python 3 distribution: Anaconda
• Other needed Python package name(s): pillow

(A raw image 1.tif and its fat tagged version 1_tagged.png from the directiory Pancreas_samples will be referred to demonstrate fat analysis)

1. Once this repository is downloaded in a local machine, open the Fatquant folder using an IDE that supports Python 3 (e.g. Visual Studio Code).
2. Run the file named tool_dos.py under an interpreter of Anaconda and execute options from the menu as per requirements.

This option allows binary thresholding to be performed on either a raw image or a tagged image in each attempt.

For raw image, users can refer the following inputs:

Enter threshold value: Pancreas_samples/1.tif

Enter threshold value: 230

For image with manual tagged fats, users can refer the following inputs:

Enter image name with extension: Pancreas_samples/1_tagged.png

Enter 'y' if you want to use default manually tagged color value (R,G,B: 255,255,0): n

Enter value for Red channel: 255

Enter value for Green channel: 255

Enter value for Blue channel: 0

(For this image, default color option would have given the output but manual entry of color was chosen to demonstrate what values to enter)

This option allows segmentation of white pixels to be performed on binary thresholded raw image or manually tagged image.

For raw image, users can refer the following inputs:

Enter thresholded image name with extension: Pancreas_samples/1_230.tif

Enter 'y' if the thresholded image is of manually tagged data: n

For image with manual tagged fats, users can refer the following inputs:

Enter thresholded image name with extension: Pancreas_samples/1_tagged_bi.png

Enter 'y' if the thresholded image is of manually tagged data: y

This option allows quantification of fats from white segments generated through Option 2.

Users can refer the following inputs:

Enter original image name with extension: Pancreas_samples/1.tif

Enter thresholded image name with extension: Pancreas_samples/1_230.tif

Enter output image name (exclude exrtension): Pancreas_samples/1_230

Enter 'y' for default white group names: n

Enter white groups combined name (extension 'csv'): white_groups_combined.csv

Enter white groups segmented name (extension 'csv'): white_groups_segmented.csv

Default color value: machine tagged value (R,G,B: 0,255,255)

Enter 'y' if you want to use this value: n

Enter machine tagged value for Red channel: 0

Enter machine tagged value for Green channel: 255

Enter machine tagged value for Blue channel: 255

Enter minimum fat diameter: 27

Enter maximum fat diameter: 130

A file named 1_230_27_130.csv will be generated which stores the total area (in pixels) identified as fats by Fatquant.

(Default options for white groups name and color value would have given the output but manual entry was chosen to demonstrate what to enter)

This option allows comparison of machine identifed fats (generated through Option 3) with manual tagged fats.

Users can refer the following inputs:

Enter machine tagged image name with extension: Pancreas_samples/1_230_27_130.tif

Enter manual tagged image name with extension: Pancreas_samples/1_tagged.png

Enter 'y' for default white group names: n

Enter file name for manually tagged white groups combined (extension 'csv'): white_groups_combined_manual.csv

Enter file name for manually tagged white groups segmented (extension 'csv'): white_groups_segmented_manual.csv

Enter file name for machine tagged fat areas (extension 'csv'): Pancreas_samples/1_230_27_130.csv

Default color value: machine tagged value (R,G,B: 0,255,255)

Enter 'y' if you want to use this value: y

Enter manually tagged value for Red channel: 255

Enter manually tagged value for Green channel: 255

Enter manually tagged value for Blue channel: 0

Enter intersection value for Red channel: 127

Enter intersection value for Green channel: 255

Enter intersection value for Blue channel: 127

(Default options for white groups name would have given the output but manual entry was chosen to demonstrate what to enter)

There are three programs included in the extras folder of this repository which are not needed for running Fatquant tool but their usages are mentioned below.

1. fat_results_compare_others.py is used to compare image outputs generated by tools other than Fatquant with manual tagged image.
2. white_segments_uncombined.py is used to get a view of white segments where the tiles are not combined. File tile_size_350_sample.png is an output generated with this program where the parameter value of s_size (tile size) was set to 350.