Dear Kristine,
I am currently experimenting with prepending PreFreeSurferPipeline.sh to recon-all (of HCP pipelines https://github.com/Washington-University/Pipelines ) which does APCP alignment, robust FOV (cropping of unrelevant parts of image FOV - neck etc.), bias correction and (optionally) FNIRT-based skullstrip and gradient nonlinearity correction.
I believe that ACPC alignment could prevent Talairach registration errors/inaccuracies in recon-all which could have profound effect on subsequent skullstrip and subcortical labeling (it could at least prevent to disastrous errors which some other users reported recently, such as this: http://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg51789.html). This could also help to mutual registration of data of multiple sessions (i.e. in longitudinal studies) to compensate different head positioning.
It could be also beneficial to use FNIRT-based skullstrip to help (or even to replace) freesurfer's skullstrip (as is it done in FreeSurferPipeline.sh of HCP pipelines), since, to my experience, the inaccurate skullstrip is basis of most errors in FreeSurfer reconstruction stream. This FNIRT skullstrip works fine with our high-quality data from 64-channel coil acquired by protocol similar to Human connectome project (I ocasionally find only several, probably negligible amount of voxels of gray matter cut-out), but not as good with older data where significant parts of gray matter are often cut-out.
The drawback is that when you want to get to the original space, you have to account one more transformation and there is one more step in processing pipeline you need to check.
This is my view based on current experience. I would also greatly appreciate any comment (different perspective) of other users.
Regards,
Antonin
Dear Freesurfer team,
This may sound like a bit of a small question, but I could not find any exact answer elsewhere.Does the exact orientation of the input image have much effect on the accuracy of the segmentation in Freesurfer? It seems that some people apply ACPC alignment to the output of Freesurfer, so I am confused as to which order would be optimal.
Thank you in advance, Kristine
freesurfer@nmr.mgh.harvard.edu