[Trackmate] x,y coordinate standard deviations

imagej
trackmate

#1

Hi all,

Is there a way to get out the standard deviations for the x and y coordinates for the detected spots in Trackmate?

Thanks!

EDIT: To clarify, I’m looking for the standard deviation in the determination of the coordinate, not the standard deviation in the brightness values - i.e. the uncertainty in determining the x and y coordinates for each spot in each frame.


#2

Other than pick up the results per TRACK_ID for x and y from the Trackmate results table with getValue() and Array.getStatistics(array, min, max, mean, stdDev)?


#3

Yes, other than that.

From what I understand the min, max, mean, and stdDev values give those statistics for the pixel values in the detected spot (as per https://imagej.net/TrackMate_Algorithms#Mean.2C_Median.2C_Min.2C_Max.2C_Total_intensity_and_its_Standard_Deviation).

I’m interested in getting out the uncertainty in localizing the spot initially - i.e. the standard deviation in the position. I’ve gotten this metric out from other algorithms before, so I was hoping there was a way to do so here as well.


#4

The Array.getStatistics() gives the statistics of the ‘array’ argument. So if you throw it an array that contains x values, it will calculate the stdDev of the x-coordinate values. Same for an array containing y-coordinates.
How to put in the x- and y coordinates, and how to have a StdDev of them, is up to you. Wouldn’t you want to have a series of locations (x,y) and calculate the StdDev of that?


#5

No, that’s not what I’m looking for.

Let me try to clarify:

I’m looking for the localization noise for each point. This would probably be linked to the particle detection in each frame. Essentially this is the uncertainty in the determination of the coordinates of each point.

I’m using this value in calculating diffusion coefficients from mean-square displacement values using the following equation:

MSD(τ) = 4Dτ + 4σ^2, where MSD is mean-square displacement, τ is the time-lag, D is the diffusion coefficient, and σ is the localization noise that I’m looking for.

This is important for my purposes because the diffusion of the particles that I’m looking at can be on the same length scale as the typical localization error for most SPT algorithms, so to accurately calculate D, it’s necessary to know σ.
(This concept might be better explained in this paper: https://journals.aps.org/pre/abstract/10.1103/PhysRevE.82.041914)

I was previously able to get this value out for all of the detected points when using the MATLAB implementation of u-Track, but I wanted to switch to TrackMate, and I was hoping to get out the same parameters here.

I hope that this somewhat clarifies what I’m looking for, and thanks for taking the time to respond so far!