 |
 |
Adding to Stefano's question, aren't FA values and connectivity values independent (i.e. the number of samples in a voxel that are parallel to the previous voxel's connectivity results has nothing to do with either voxel's FA value)? If this is the case, and if your goal is to make claims about differences between subjects or between groups for FA values, it seems best to use TBSS. However, if you are interested in drawing conclusions about variability in a specific pathway found by tractography, can you instead do a comparison of the connectivity distributions? For example in Stefano's case, divide the connectivity values at each voxel by 5000 and compare the resulting distribution to tractography results from other subjects' thalamus/ cortical ROIs. Thanks! ted On Mar 6, 2008, at 4:33 AM, Saad Jbabdi wrote: > Hi Stefano, > > I think that you'd better go with the first solution, i.e. average > FA along the tract with no weighting. > Weighting FA with the probabilities assumes at least two things: > the probabilities represent the amount of the tract in each voxel > relative to other tracts, and second that FA is a measure that is > specifically related to your tract. None of these is correct, in my > opinion. > So the best thing to do is make sure you're looking/dissecting a > tract that makes anatomical sense (such as you did with your target > cortical mask), and average FA along this tract. You might also > want to avoid including crossing fibre voxels, which would avoid > biasing your estimate of the average FA toward low values that are > only due to partial voluming. > > Cheers, > Saad. > > > On 6 Mar 2008, at 04:31, Marenco, Stefano (NIH/NIMH) [E] wrote: > >> Hi, I have a question regarding how to calculate FA for a tract. >> I send 5000 fibers from a roi the thalamus to a certain cortical >> mask. >> I obtain a tract with intensity values ranging from 1 to 22000. >> I can average the FA over any voxel belonging to the tract, but this >> weights voxels with very low probability of belonging to the tract >> the >> same as those that are in the middle of white matter, so I think >> this is >> wrong. >> I can define the tract as any voxel with at least 50% probability of >> connection to any voxel of my seed region, i.e. any voxel with at >> least >> 2500 fibers passing through. Do the FSL experts agree on this type of >> threshold? This would be a "hard" segmentation of the tract. >> >> I also can weight the FA at each voxel by the intensity of the tract >> image (a "soft" segmentation). In this case, though, I think that a >> ceiling should be imposed, i.e. that any voxel with a tract >> intensity > >> 5000 should be counted only once, while a colleague of mine thinks >> that >> the proportionality should be maintained regardless of the nominal >> tract >> intensity, i.e. that a voxel with 22000 fibers passing through >> should be >> weighted 4.4 times one with 5000 fibers passing through. What is your >> advice in this case? >> >> Thanks for your help in advance. Stefano Marenco >> > > ---------------------------------------------------------------------- > ----- > Saad Jbabdi, > Postdoctoral Research Assistant, > Oxford University FMRIB Centre > > FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK > +44 (0) 1865 222545 (fax 222717) > [log in to unmask] http://www.fmrib.ox.ac.uk/~saad > ---------------------------------------------------------------------- > ----- > > > >