Tractography measures

[Ehsan Misaghi]

Hi everyone,

We have analyzed a dataset using tractography in ExploreDTI and we have extracted the tract information. There are a few questions I have about the spreadsheet when we are converting the mat files of the tracts into a text file:

1. Mean tract length seems to be (((number of points)/(number of tracts))-1). So, does this mean that the number of points is simply the number of voxels that the tracts are passing through? Why that -1 is included in the calculation?

2. How the approximate tract volume is computed? I thought that it is the number of voxels multiplied by the voxel volume, but I calculated those (using points = voxels!) and figured out that it's not!

3. Is the mean angle the average of the angles each glyph makes with the next one in each tract averaged for all the tracts?

4. MD is (lambda1+lambda2+lambda3)/3 and RD is (lambda2+lambda3)/2. One can calculate each of these separately by having the values for the mean lambdas. Is there a specific reason why RD is included in the spreadsheet, but not MD (or AD), though they both can be calculated from the lambdas?

Thanks,

Ehsan

[Ehsan Misaghi]

And of course one other question would be: how approximate tract volume measured from DTI tractography compares to the regional white matter volume measured using VBM on the T1 image?

Thanks,

Ehsan

[Sjoerd Vos]

Hi Ehsan,

Concerning your initial post:
1) The number of points is not necessarily the number of voxels it passed through. In the Tractography settings you can define a step size, which defines how big a distance (default 1mm I think) is travelled from one point along the tract to the next following the first eigenvector. The length of a tract is thus it's (number of points - 1) times the step size. the mean tract length is simply the average length of all tracts
2) correct, it is the number of voxels that the tract bundle passed through times voxel volume
3) I believe so
4) The MD is also in the table (or was at least in v483), but denoted as ADC.

I'd say there's no easy comparability between tract volumes and VBM, given the inherent difference in imaging these are based on.

HTH,
Sjoerd

[Lara Schlaffke]

Hi Sjoerd,
I´ve tried to compare approximate tractvolume (of muscle tracts) with volumes generated manually based on T1 weighted images (conventionally volumetry). Unfortunately, there are enourmus differences in the volumes (up to power of 10). Although tractography works pretty well and represents muscles of interest...
Do you know, whether certain voxels maybe counted multiple times, if multiple fibers passed through for the approximate TractVolume, or do you have another idea where this might come from?

Best,
Lara

[Sjoerd Vos]

Hi Lara,

The tract volume you get is just the number of voxels visited by at least a tract multiplied by the voxel volume. I agree it's likely that your dMRI resolution is lower than the T1, so increased partial voluming could slightly overestimate tract volume, but such a huge overestimation seems very strange. You don't have an error in one of your voxel sizes by any chance? I know it's unlikely but your results seems very weird.

Sjoerd

[Lara Schlaffke]

Hi Sjoerd,
I just figured, that the HeaderInformation of the T1 voxelsize were misleadingly set to the voxelsize of the dwis although the images were not changed..... I don´t know how this could happend, but this might explain the huge differences.
Thank you very much for that tip
Lara


PS: Ignore my previous post :)

[Cristina Simón Martínez]

Dear ExploreDTI experts, 

I am running a CSD-based tractography in a dataset to analyze de corticospinal tract. When I click on the stats to see the values I see that the N-value is very high. I thought this would be the number of tracts, but according to the literature I find it to be very high. In the population I am studying, the fiber count should be around 2500-3000, but the N-value says 20115. Could anyone confirm that the number of fibers is actually this N-value, please? If not, how could I get the actual number of fibers?

Thanks a lot in advance!

Regards, 

Cristina