Hi All,
I performed tbss steps 1-4 on a group of subjects using the -T target and all seemed to work fine. I checked the all_FA image against the mean skeleton and all the images were reasonably well aligned with the skeleton.
I then decided I wanted to register to a study-specific target, mostly since the sample I am working with is older (mean age of 72). I first ran tbss_2_reg -n within the control group (n = 28) to identify the best target for intersubject alignment. I went ahead and completed steps 3-4 and when I check the registration most of the alignments were good except for the final volume in the series. It looked like the volume was pulled down (inferior) in the coronal plane such that the corpus callosum skeleton was well above the CC on the registered FA map. I went ahead with my plans and used the same identified volume as a now predetermined target for the whole sample (n = 56), hoping that, somehow against all sense, this issue would resolve itself. Alas, it did not and that lone subject is misaligned with the skeleton. I examined the FA values in the misaligned portion of the suspect volume in all_FA_skeletonized and the values were far below the preceding volumes, doubtless due to the misalignment. Some skeleton voxels had values below or hovering about .20, despite setting the threshold to that level.
So, I could just stick with the initial registration results (those to the FSL-provided target), but I'd like to figure out what has gone wrong with that one volume. Why has it registered fine to the FSL-target, but not to the selected as the best match. Perhaps I'm misunderstanding the mechanisms of the -n option.
Some other information: the volume (all, in fact) is in radiological orientation and it appears to display fine on fslview. Is there something else I could check about this volume that may explain the misalignment?
Thanks for your help,
-S
|