has anyone on here successfully managed to use the 2nd part of the LAP tracker to assign the same label to mitotic daugthers and their parent?
I am looking at HeLa cells and managed to optimize the first part of LAP tracker quite efficiently, and have hardly gaps in segmentation, but am really stuck on the mitosis aspect which I need!
- HeLa cell ~ 120px diameter (ish) in interphase (~ellipse)
- HeLa mitotic: ~120-150px in length, 50px “height” (~rectangle)
- pixel size 0.120micron
- frame to frame displacement max displacement interphase: ~2-40px (if cells make a big jump)
- mitosis displacement (when parent-1daughter tracked): ~50-100px (~60 displacement if get centroid to centroid)
- max gap frame or 2 frames at most due to mis-segmetnation
My 1st phase optimization settings were: 20sigma, 20-120px search radius works really well and I can follow all interphase cells and sometime connect 1 parent to 1 daughter cell after mitosis.
My 2nd phase settings: I don’t want to have merges or splits so I have the following settings:
- split alternative cost = 1
- merge alternative cost to 1.
- gap closing cost: 40 (displacement of the object btw frames)
- mitosis alternative cost: 80 (roughtly the cost of closing 2 gaps)
- max gap displacement in px: 40 (about the max displacement of a cell if being mis-segmented)
- max merge score 1 (coz I don’t want any merges)
- max split score 1 (coz I don’t want any splits, just mitoses)
- max temporal gap: 2 frames
- max mitosis distance, in px: 60 px (centroid to centroid distance daughter to parent in previous frame)
and… it doesnt work
Any insight /examples that work would be much appreciated so I can understand a bit better how LAP wokrs (though I am digging deep into the original code on github).
Output-1stphaseOnly.tif (7.7 MB)
Output-mit2ndphaseLAP.tif (7.7 MB)