Segmentation code for the Cellavista and BD pathway imaging platforms

Version 2.17.16

Christian Meyer (christian.t.meyer@vanderbilt.edu)

Currently under construction

• Integration of Bayes Classifier
• Tracking Code
• Making GUI windows normalized to the screen they are on
• Compile in a matlab project to run on other computers without matlab
• Put a button to open SegmenterV2 GUI on the FileSorterGUI
• Put widget to open bayes classifier from SegmenterV2
• Generate introduction box when first initiated to give run down of pipeline
• Put widget in Bayes classifier gui to move to next image
• Put widget in bayes classifier gui to build logistic regression classifier
• Put widget in bayes classifier to run cell tracking algorithm based

Description

Code works by calling FileSorterGUI first from the command line to sort the images in the experimental folder generated from the Cellavista After closing the Window, Run SegmenterV2 from the command line to open the GUI involved in segmenting the cells.

Algorithm follows the following:

1. Illumination correction by either cidre correction where a predefined CIDRE map has been generated or by subtraction of a control image.
2. Otsu's multithresholding algorithm to binarize image
3. Filter image with matlab's imtophat function to remove noise
4. Water shed segmentation by image inversion -> pixel suppression -> and watershed segmentation using matlab function
5. K-nearest neighbor algorithm to predict cytoplasm

Segmentation can be run in parallel. Currently the waitbar assumes the number of available workers is 4.

Next is creation of a classifier to generate probabilities associated with the characteristics of the nuclei generated from the segmentation used in the subsequent tracking algorithm