nkorobkov / microglia-or

Scripts to segment microglia images and generate reports

microglia segmentation

Set of algorithms to segment microglia cells from mouse brain shots. (There is a ton of science behind how those pictures are taken. This repo has nothing to do with that.)

Algorithm receives raw images as input and finds microglias borders and hull convex for each cell.

All data is reported in csv files generated for each image. Like:

#, Size, HullSize, Perimeter
 1, 8298, 26190.5, 1658
 2, 3544, 8829.5,  958
 3, 3378, 19413.5, 1197
 4, 3130, 11644.5,  826
 5, 1899, 9405.5,  752
...

Images themselves look shomething like this:

Unprocessed picture

Segmented picture

Publications

This software was used to produce results presented on several scientific conferences. Here is the list of some publications:

Glyavina M.M., Loginov P.A., Dudenkova V.V., Reunov D.G., Karpova A.O., Prodanets N.N., Korobkov N.A., Zhuchenko M.A., Schelchkova N.A., Mukhina I.V.
Carbamylated darbepoetin (CdEPO) impact on microglia activation in a model of occlusion of the middle cerebral artery.
"XIV International Interdisciplinary Congress "Neuroscience for Medicine and Psychology"
2018 May 30- June 10 p. 159

Glyavina M.M., Loginov P.A., Dudenkova V.V., Shirokova O.M., Reunov D.G., Karpova A.O., Prodanets N.N., Korobkov N.A., Zhuchenko M.A., Mukhina I.V.
Application of laser scanning confocal fluorescent microscopy for visualization microglia morphology in mouse local cerebral ischemia.
"26th International Conference on Advanced Laser Technologies" [ALT'18].
2018 September 9-14: B-P-2.

Glyavina M.M., Loginov P.A., Dudenkova V.V., Shirokova O.M., Reunov D.G., Karpova A.O., Prodanets N.N., Korobkov N.A., Zhuchenko M.A., Mukhina I.V.
Morphological analysis of microglia in early postischemic period in the mouse local cerebral ischemia.
"TERA — 2018, 3rd International Conference Terahertz and Microwave Radiation: Generation, Detection and Applications"
2018 October 22—25: S10, P.10.9

microglia-or's Issues

Add hull convex to report

Report hull convex area to reported info.

Need to know wether to add axons to convex or no.

https://docs.opencv.org/3.0-beta/doc/py_tutorials/py_imgproc/py_contours/py_contours_more_functions/py_contours_more_functions.html

Connection lines does not add up to the edges

Connecting two components with a line means that this line actually exists in different dimention and we need to use this info when calculating perimeter.

Consider recalc all edges after merging.

Redundant info about noise components is stored

At some point during the algorithm we decide not to use some components (possible axons without host nucl around).
We still have info about those in markers and components index. It may be useful to discard it not to iterate over it during edges and sizes recalculations.

Citation/Methodology

Hi @nkorobkov ,

Thank you for this repo. I found your notebooks super helpful in my own microglial segmentation. I have a couple of questions:

1. How do I cite this work?
2. Is the methodology explained/from some other paper that I could look more closely into?

Best,
Aditya

Connection lines is drawn between two random points

when two components connect line is drawn to illustrate connection. Line is drawn between two edge points that are closer than threshold. But it would be nice to draw it between points with minimal distance.