Stay with same cell
Opened this issue · 2 comments
Timepoint t
Timepoint t+1
As indicated by the ImageJ ROIs, the algorithm decided that the optimal solution is to assign the "wrong" cell as the next cell in the track. Correct would have been the cell on the right to the selected cell in t+1. For the human this is obvious, because the "correct cell" would have been the one with a similar shape and also continuing the movement in a similar direction.
Are there parameters that we can try to improve the tracking along those lines?
Hi @tischi, we have yet to implement a solution for this.
We are long due to adding an API for custom association scores. Would that help you?
Related issue #46
Tuning the appearance and disappearance penalization parameters could help if the neighboring cells are disappearing/appearing and not moving out of the z-plane, but I can't tell from these images.
We are long due to adding an API for custom association scores. Would that help you?
That sounds like the thing that would indeed help :-)
TrackMate has that and in fact seems to currently sometimes to a better job of following "the same" cell...