Hi Josh,
> I believe the number in the waytotal output is the streamlines
Actually, it is the total number of streamlines that made it from seed to target - given the number of streamlines you have chosen to send out.
> what I can infer from this number, (i.e. directionality).
You certainly can not infer directionality of the pathway from this metric.
> this number was essentially useless for analyses purposes
I wouldn't say "useless" but you need to be particularly cautious what to do (and not) with it.
If you divide your fdt_path(.nii.gz) by the waytotal you get the >conditional voxelwise probability< to successfully track the pathway using bedpost(x)/probtrack(x) under the prior that the tract actually exists.
This is because according to Bayes' theorem we find:
p(S→V→T | M1→P) = p(S→V→T) / p(S→T)
with:
p(S→V→T | S→T): normalized conditional probability of a voxel V to belong to the pathway S(eed)→T(arget) supposing that the tract / pathway exists (which can be established by prior evidence and/or clinically)
p(S→T): total number of samples that reach T from S (=„waytotal“) using bedpost(x)/probtrack(x)
p(S→V→T): subset of samples from S to T passing through voxel V (= the number you find in a given voxel of your fdt_path.nii.gz file)
because:
p(S→T | S→V→T) = 1, i.e. the probability that the tract exists if a voxel V belongs to it is obviously 100%.
Note that "→" is simply to denote that tracking proceeds from S(eed) to T(arget) but is NOT to infer directionality of the pathway.
Now, the tricky issue is that a) voxelwise probabilities normalized by the waytotal may >seem< rather low (something like 0.1 to 1 % for well-established tracts like the pyramidal tract is not uncommon) but quite strongly depend on b) tracking criteria and c) inter-session/-subject/-sequence/-scanner variability. In that regards, you may think of the waytotal as a measure of "general trackability" of a given tract given the method (both data acquisition and analysis) and criteria you used to reconstruct it. It is certainly hard (and will usually not be wise) to compare this metric across different pathways.
However, you can show that while perifocal edema of intra-axial brain lesions reduces the total number of samples that reach T from S (i.e., the waytotal) the conditional probability to reconstruct, for example, the pyramidal tract by identical criteria is preserved on the lesional as compared to the contralesional side when you >know< that the tract exists (e.g., from clinical evidence).
So one useful measure may be comparing the waytotal of the same tract, reconstructed by identical unbiased criteria, across hemispheres.
You may want to check out the list and Heidi's & Tim's Diffusion MRI book (incl. my chapter) for further details.
Hope this helps a bit.
Cheers,
Andreas
Am 04.01.19, 20:44 schrieb "FSL - FMRIB's Software Library im Auftrag von Josh Robinson" <[log in to unmask] im Auftrag von [log in to unmask]>:
Hello FSL experts,
I am wondering how I should interpret the output of probtrackx. I believe the number in the waytotal output is the streamlines, but I am wondering what I can infer from this number, (i.e. directionality). A colleague who went to a workshop in San Diego asked about this and said this number was essentially useless for analyses purposes. If this is the case, is the output of interest instead the mask generated?
In general I am wondering what the consensus is on how probtrackx output is intended to be analyzed.
Any guidance is much appreciated,
Josh
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1
