Skip to content

Image segmentation method for the automatic bulk detection of cells and their vacuoles

License

Notifications You must be signed in to change notification settings

Smith42/cell-vacuole-finder

Repository files navigation

Automatic detection of cells and vacuoles through image segmentation

A watershed segmentation method, with cell nuclei acting as initial markers, is used to automatically detect cells from plates scanned by the IN Cell Analyzer 6000. Morphological reconstruction by erosion is used to find dark spots in each found cell, and label them as vacuoles. Finally, a shifted interquartile range is applied to the found cells’ logarithmed sizes, to detect and remove outliers. A python script has been written to output an RGB image with all found cells and vacuoles labelled, and detected anomalies labelled in blue. An example is below. The script also outputs two log files per slide, describing attributes of each found cell and vacuole. This method takes around 5 seconds to scan a slide, and has been adapted to process slides in bulk from a given directory.

Example output

Please see the full report for more information on the algorithm used.

Usage

processDir.py finds all slide images matching a regex in a given directory, and runs the algorithm on them. genExample.ipynb runs the same algorithm interactively in a Jupyter notebook. genIQR.ipynb finds the interquartile range of a size sample, and returns the likely upper and lower size bounds of inliers.

About

Image segmentation method for the automatic bulk detection of cells and their vacuoles

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published