I'm not clear what the purpose of the thresholds and binarizing is - do you actually not care, in each subject, when a certain path is significantly more often taken than another, when both of them pass the "ideal" threshold? If the idea is to weight subjects equally (note that binarizing won't actually make this the case, because different subjects could have significantly different numbers of above-threshold voxels), then that could probably be done with a normalization rather than a thresholding. Unthresholded data has more information in it, by definition.
However, I haven't done any tractography analysis personally, so this is mostly speculation. Tim On Mon, Jul 30, 2018 at 7:46 AM, Amin Dadashi <[email protected] > wrote: > Dear HCP experts, > > I want to perform a seed-based (preferably probabilistic) tractography for > some brainstem ROIs using the HCP 7T DTI dataset. > > To do so, I used FDT probtrackx2 for every subject to calculate the ROIs > streamlines. In order to do a group analysis (replicating the method here > <http://dbic.dartmouth.edu/wiki/index.php/Diffusion_Tensor_Imaging_Analysis#Combining_Individuals_into_a_Group>), > I created a binary pathways mask by puting the threshold#1 on each > individual probabilistic map "fdt_paths.nii.gz". Then, I chose the > threshold#2 for the summation of the individuals' binarized connectivity > map to find the most recurring streamlines among subjects. > > The value of each "fdt_paths.nii.gz" voxel shows the accumulative number > of streamlines from all the voxels under the ROI mask. > > Now, the issue is how to choose threshold#1, which is probably dependent > on many factors; specifically the tracts length (and also maybe seeds > location and volume?...). And a more basic question is if this whole > approach makes sense at all, since no inferential statistics is being used? > > Thank you in advance. > > Best, > Amin > > _______________________________________________ > HCP-Users mailing list > [email protected] > http://lists.humanconnectome.org/mailman/listinfo/hcp-users > _______________________________________________ HCP-Users mailing list [email protected] http://lists.humanconnectome.org/mailman/listinfo/hcp-users
