TrackMate Which Pixels are Used to Calculate Mean Pixel Intensity

Hello all,

I’m using TrackMate to capture fluorescence data from a z-stack of vesicles ranging in size from 5-20um in diameter. I’m wondering how TrackMate calculates mean pixel intensity.

First, below are three points from http://imagej.net/TrackMate_Algorithms. I think these say that the mean pixel intensity is measured within R? Given the range of vesicle sizes I generally set R to be 11um to capture the range of sizes, but if R is used to calculate mean pixel intensity I would also be averaging pixels outside the smaller vesicles.

  • "R the spot radius, also in physical units. The current detectors only set this radius value to be the one specified by the user."
    
  • "Estimated diameter This feature estimates an optimal diameter for each spot, based on contrast calculation."
    
  • "Mean, Median, Min, Max, Total intensity and its Standard Deviation The plain statistical estimates are simply calculated from all the values for pixels within the physical radius from the spot center."
    

Second, I’m tracking vesicles in a z-stack. Does TrackMate find spots in every z slice? Are these spots linked in the z dimension to get a spherical volume? And finally, if volumes are identified, are all the pixels in the volume used to calculate mean pixel intensity?

Thank you for you time,

pcaveney

Hi @pcaveney

1.

Yes, the mean, median, etc… are all calculated with the radius R. It defines a ROI in the shape of a disc in 2D and of a sphere in 3D.
So you are also right: if your actual vesicles have a smaller radius than R, some pixel values outside of the vesicles will be included in the mean calculation.

2.

Euh sorry I did not het that question. Here is a try:
TrackMate works in 3D, and will find vesicles in a Z-stack as well.

The ROI for every spot is indeed a sphere of radius R in physical coordinates. If the z-calibration is different than the x & y calibration (for instance pixel width = pixel depth = 0.2µm but voxel depth = 0.5µm), then the ROI is still a sphere in physical coordinates. If R = 1µm then all the pixels within a physical distance of 1µm will be included in the mean, that is, using our previous example:

  • 5 pixels away from the center in the XY equatorial plane.
  • 2 pixels away along Z.
2 Likes

@tinevez, thank you so much this answers my question! I’ll try either filtering only vesicles larger than R or, for the smaller vesicles, only taking the brightest 200 or so pixels in R and averaging those.

1 Like