 |
 |
Hi, are you tracking in diffusion (B0?) or native highres (T1?) or MNI space? I would make sure that the preprocessing (eddy current correction, in particular) and registration looks good. Then, yes - I think using the CC as a waypoint and possibly the ventricles as an exclusion mask would be a safe choice. In general, you should not hesitate to make use of your anatomical prior knowledge to increase the specificity of your results. Cheers, Andreas Am 05.07.2013 um 23:43 schrieb Anna Greenwald <[log in to unmask]>: > Dear experts, > > I guess the subject line is just stating the obvious; the problem is really that I'm unsure which constraints to use and how much to trust my results. Here's the issue: > I'm trying to localize the fiber tracts connecting left and right auditory cortex via the corpus callosum in order to then be able to compute stats on those connections. > > My first approach with probtrackx2 was to simply track from one cortex (as the seed) to the other (as the waypoint, target, and also stop condition); do that for both directions, and combine the results. When I looked at the identified tracts, they connected through the posterior part of the corpus callosum, which is good, but on their way there, they made a shortcut through the ventricles in some cases, which is clearly nonsense. > > So my second approach used a "high MD / CSF" exclusion mask to prevent that. As a consequence, the fibers no longer ran through the ventricles, but in some cases didn't cross the midline through the corpus callosum anymore, but superior to it (and in some cases, through fairly low-FA regions). > > I don't really want to use an "let fibers only go through high-FA regions" constraint, because low FA doesn't necessarily mean that there isn't a strong fiber connection in a certain direction - it could just mean that there are equally strong connections in orthogonal directions. > > I'm aware that I could set the corpus callosum as a waypoint to force the fibers to go through there, but considering that they always seem to take the shortest possible way given the constraints (i.e. not going all the way down into CC when forced to avoid the ventricles - and thus climb higher, and simply going straight through the ventricles when connecting through the CC), I'm not sure how much I could trust the results. Sure, if I force the tracking to avoid the ventricles and to go through the CC, the connections will look plausible - but that may simply be because I've forced the fibers to connect my two ROIs via the CC and at the same time prevented them from going anywhere implausible, not because the fibers actually run there. Even if the true connections ran through anterior CC (which I know they don't, I'm just using this to illustrate my point), I might see them connect through posterior CC simply because probtrackx2 finds the shortest connection given my constraints. > > Do you have any suggestions on which constraints are "safe" to use and how I can assess whether my results are indeed valid? I know that anything connecting left and right auditory cortex has to cross way stronger anterior-posterior tracts along the way, so it's tricky, but I was hoping that probabilistic fiber tracking could do it. Maybe that was overly optimistic, or maybe I'm just not doing it right? Any input is very much appreciated! > > Thanks, > Anna