TrackMate questions

script
tracking
imagej
plugin
segmentation
trackmate

#1

I have several questions regarding the TrackMate plugin for FIJI.

I’m interested in tracking cells over time using one stack (which is a raw image in a 16-bit format) and then calculating their features using a different stack (which is in 32-bit format and contains NaNs). The 32-bit stack is a ratio between two channels, after background subtraction and thresholding.
The plugin works well on the 16-bit stack but it can’t handle the NaNs in the 32-bit stack. Also, because it’s a ratio image the cell detection is biased towards higher ratios.
In addition, the cells I’m tracking are not round, and for the measurements (such as mean intensity over time) I prefer to use my own segmentation script (unless there is an option to track objects with different shapes).

  1. Is it possible to modify the plugin so that it will work with a stack that contains NaNs?

  2. Is it possible to export the tracks and spots calculated from one stack using the TrackMate plugin to another stack, so that I could measure the spot properties using a different stack?
    I was only able to copy the overlay or to get one ROI containing all the tracks together.
    More specifically, I would like to use TrackMate within a script on one stack to track the cells, and then export the track IDs and assign them to objects I detected using my own segmentation on a different stack.

  3. In addition, is it possible to modify the plugin so that the shape of the spots will be more flexible, for example add a circularity feature (similar to the same feature in the “analyze particles” function) that can be modified? That way, objects could be segmented according to their borders (and not as spots), and the measurements could be done directly on the detected objects using the plugin.

Thanks in advance,
Avia


#2

Hi Avia.

Unfortunately I mainly have bad news.

$1.
TrackMate won’t be working well with NaNs. If you sum a number with a NaN, the result is a NaN. So all neighborhood operations will be ‘contaminated’ by NaNs. This means that almost all spot analyzers and spot detectors will fail. You have to remove the NaNs from the stack, but to do so you have to put a value that makes sense. For instance, replacing it with 0s might not be correct scientifically, depending on how you generate the stack.

Another approach would be to write a Spot analyzer that takes into account NaN and skips them. I refer you to the documentation.

$2.
Yes this is possible, but you have to manually edit the TrackMate XML file. Open it in a text editor. The XML file is organized hierarchically be sections, using tabs in in <>. Go to the Settings > ImageData section and edit the filename and folder to point to the image file you want to analyze.

The load this TrackMate file. It should load the image you pointed to. In the last panel of TrackMate, there is an action called Recalculate all features. This will recalculate all features of TrackMate using the new image.

$3.
No it is not possible at the time. TrackMate is limited to handling spots that are basically a X, Y, Z coordinates with a radius.


#3

Bonjour,

Tout d’abord je tenais à vous féliciter et à vous remercier pour ce super logiciel, il facilite grandement la tâche !
Par contre, j’ai eu quelques problèmes concernant le reloading d’un trackfile. Lorsque je charge celui-ci, trackmate m’ouvre le film associé au tracking avec les tracks (jusque là pas de problème). Toutefois, lorsque j’avance dans le film, je perds le suivi des points (dots) associés à la track, ceux-ci deviennent roses et ne bougent pas. De plus, plus moyen de suivre les points de la track via tracksheme.

Existerait-il une solution à ce problème ?

Merci d’avance,
Francois T.