How to prevent loss of cell tracking in highly dense samples?

[Erwin-Jo]

Hi,
I wanted to ask if you have any advice on handling biological samples with very high cell density. The dense packing often leads to incorrect label assignments between neighboring cells, causing the tracking to break.

Additionally, due to the three-dimensional nature of the sample, some cells disappear from the image frame and reappear later, which also results in loss of tracking.

So far, I’ve tried adjusting the parameters by increasing max_neighbors and decreasing both distance_weight and max_area, but I’m still encountering issues.

Any suggestions would be greatly appreciated

[JoOkuma]

Hey @Erwin-Jo, how motile are your cells?
The cell density often induces segmentation mistakes, but not tracking assignment issues.

For the reappearing cells, I recommend using the close_track_gaps function to connect them as a post-processing step.

[Erwin-Jo]

Thanks for answering so quickly. I will definitely try the close_track_gaps function.

Regarding cell motility, I would say they are not very mobile. First, the cell type(s) I am dealing with are not known for their motility. And secondly, the high packing density would make jamming each other more likely.

Also, I realized that I phrased the "disappearing" & "reappearing" not precise enough. I think a better way to describe it might be this.

Let us assume one cell gets the label i assigned on the initial frame A and then later the same label i is assigned on a later frame B. The issue is that it is very unlikely that the original cell from frame A with the label i moved so fast or far away in the time from A to B (where we did not observe it) to be same cell as in frame A. It is more likely that the same labeling i was reused for a different cell.