 |
 |
Thank you Saad for your reply. I know questions like these are a frequent topic on the forum and I appreciate you taking the time to address them. For this study, the seeds and targets were derived from individual Freesurfer cortical and subcortical segmentation and registered to the DTI space. So, unfortunately for me, the seeds and target ROIs won't be equal across subjects. Maybe, I'm doing/understanding something incorrectly here. I used one seed and one target, with the target serving as a waypoint and exclusion mask as well. I was curious to see the streamline count for the number of streamlines going from my seed to target. When I calculated that streamline count (via fslstats seeds_to_target -M and multiplied by the number of seed voxels ), I noticed that this number was exactly the same as the waytotal. I suppose in this way that is not surprising that the waytotal, in that set up, is related to the seed_to_target. It, the waytotal, just doesn't seem to be the appropriate control to normalize the values of seed_to_target, with this setup. Taking the mean of seeds_to_target/waytotal I essentially end up with the inverse of the voxel count: $ fslstats seeds_to_OFC_LEFT.nii.gz -M 1960.647887 $ fslstats seeds_to_OFC_LEFT -V 71 748.828125 1960.647887 X 71 = 139205.99 $ more waytotal 139206 $ fslmaths seeds_to_OFC_LEFT -div `cat waytotal` seeds_to_OFC_LEFT_div_waytotal $ fslstats seeds_to_OFC_LEFT_div_waytotal.nii.gz -M 0.014085 1 / 71 (seed voxel count) = 0.014085 So, when using probtrackx2 to track between only two regions [A to B (with B as a waypoint and termination mask)] how best can I "normalize" the seeds_to_target values, since, with that specific set up, it seems the waytotal is not appropriate (I would just end up with the voxel count). I considered dividing by the voxel count of the seed + target, but I'm not sure if that is an appropriate method. If my question is "what proportion of sent streamlines from this seed reach this target?", is normalization still necessary? I suppose I could use the waytotal normalized fdt_paths to address this question, as well. THANK YOU again for your help and for developing these great neuroimaging tools. On Sat, Mar 22, 2014 at 6:13 AM, Saad Jbabdi <[log in to unmask]>wrote: > Hi > waytotal simply tells you how many samples were *not* rejected, so > dividing by waytotal ensures that when you are comparing different > subjects, the probability values are conditional on your > exclusion/inclusion masks. > If you have no exclusion/inclusion masks, and you use the same seed masks > across subjects, then it doesn't matter if you divide your results by > waytotal since that means dividing by the same value for all your subjects, > which will not change your cross-subjects comparisons. > > Cheers > Saad > > On 21 Mar 2014, at 14:25, P <[log in to unmask]> wrote: > > Hello All, > > I used probtrackx2 to track between A and B (with B as a waypoint and > termination mask) and calculated the seeds_to_target (B) file. I wish to > compare the connectivity between groups but I wonder how best to handle any > necessary normalization to allow for equivalent across-subject comparisons > on the connectivity values (i.e., mean of the seeds_to_target). I read that > using the waytotal may be a good approach. > > https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1106&L=fsl&P=R50105&1=fsl&9=A&I=-3&J=on&K=2&d=No+Match%3BMatch%3BMatches&z=4 > > https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1205&L=fsl&P=R7773&1=fsl&9=A&I=-3&J=on&d=No+Match%3BMatch%3BMatches&z=4 > However, in this case, where I have one seed and one classification > target, the waytotal = the mean of seeds_to_target x the number of seed > voxels. In that way, it doesn't seem appropriate to divide the > seeds_to_target value by the way total since that ends up being (1/# > voxels), when averaged across the ROI (A). > When I use more the one classification target, the waytotal no longer has > that relationship, which may be why this approach was suggested previously > (or maybe I am missing something here). > If I divide the mean of seeds_to_target by 5000 then I get the % of > streamlines from the seed that reach the target, correct? This seems to me > to be a possible measure to use for cross group comparisons. > Any advice on this topic is very welcome, > P > > > -- > Saad Jbabdi > University of Oxford, FMRIB Centre > > JR Hospital, Headington, OX3 9DU, UK > (+44)1865-222466 (fax 717) > www.ndcn.ox.ac.uk/team/researchers/saad-jbabdi > >