Analyzing TrackMate results with MATLAB
TrackMate was originally designed solely as a tracking tool. Its goal was just to generate the tracks then another tool was supposed to analyze these tracks. Nonetheless, we built some crude analysis facilities in TrackMate for rapid inspection. But in no way they can compete with a proper dedicated, custom analysis tool.
I favor MATLAB as a tool for data analysis. Therefore the interoperability facilities first deal with this software.
Location and installation of import functions
These tools take the form of several, hopefully well documented MATLAB functions that are shipped with in your Fiji installation. All of them are located in the scripts folder of your Fiji package
tinevez@primevere:~/Development/fiji> cd ~/Applications/Fiji/Fiji.app/scripts/ tinevez@primevere:~/Applications/Fiji/Fiji.app/scripts> ls InstallJava3D.m Matlab3DViewerIntroduction.m IsJava3DInstalled.m Miji.m Matlab3DViewerDemo_1.m Miji_Test.m Matlab3DViewerDemo_2.m bfopen.m Matlab3DViewerDemo_3.m importTrackMateTracks.m tinevez@primevere:~/Applications/Fiji/Fiji.app/scripts>
In MATLAB, select in the menu and add the scripts folder there (on Mac, the file chooser doesn't let you choose directories within .app packages, so you have to use the › MATLAB command
addpath('/Applications/Fiji.app/scripts') or whatever folder works for you) .
Importing track files
TrackMate has an action to export a simplified XML file containing only the tracks as spot positions. It is only suitable for linear tracks: tracks that are non-branching, non-merging. In that case, there is a MATLAB function that can use this XML file to import the tracks and retrieve the relevant metadata:
>> [tracks, md] = importTrackMateTracks('RealPhotobac20min_Tracks.xml'); >> tracks tracks = [125x4 double] [ 99x4 double] [125x4 double] [126x4 double] ... >> md md = spaceUnits: 'pixel' timeUnits: 's' frameInterval: 5 date: 'Thu, 8 Aug 2013 13:33:23' source: 'TrackMate v2.1.0'
Check the help section to learn about the output format (
doc importTrackMateTracks). But basically, each track is represented by a
double array in a cell array (one cell per particle). The
double array itself is organized as
T X Y Z with T by default being the frame integer number.
There exists one flag that can remove the Z dimension in 2D case. And another that can scale the time value so that it is in physical units rather than frame number. Again, check the function documentation for details.
There are some specialized tools in MATLAB that can exploit TrackMate results. For instance, here is a MATLAB class that performs mean-square displacement analysis. It is hopefully well documented in this MATLAB tutorial.