open-discussion > Freesurfer Cerebellum segmentation
Showing 1-2 of 2 posts
Jul 6, 2020 02:07 PM | Álvaro Deleglise
Freesurfer Cerebellum segmentation
Hi! I was wondering how can I run a segmentation of the cerebellum
with a different strategy than the one provided by the default
subcortical segmentation (which only differentiates lh and rh white
and grey matter). I'm currently analyzing longitudinal data from a
motor learning experiment in which the cerebellum it's very
important, so I would like to have more cerebellar ROIs.
I recently found an article by Makris et al., 2003, Human Cerebellum- Surface-Assisted Cortical Parcellation and Volumetry with Magnetic Resonance Imaging, but I'm not sure if there is any implementation of this strategy. Does anyone knows how can I use it? Or if there is another segmentation strategies available?
Thanks in advance
I recently found an article by Makris et al., 2003, Human Cerebellum- Surface-Assisted Cortical Parcellation and Volumetry with Magnetic Resonance Imaging, but I'm not sure if there is any implementation of this strategy. Does anyone knows how can I use it? Or if there is another segmentation strategies available?
Thanks in advance
Jul 6, 2020 03:07 PM | Paul Camacho - University of Illinois at Urbana-Champaign
RE: Freesurfer Cerebellum segmentation
Outside of Freesurfer, one atlas that is used for parcellations is
the SUIT atlas from Diedrichsen and colleagues: http://www.diedrichsenlab.org/imaging/su... I bring this atlas into subject T1 space using MATLAB and SPM for
each subject and then the T1 space version can be registered to
other spaces for use in cerebellar parcellation. In some lower
resolution spaces, there will likely be overlap with some of the
lobule masks and the Freesurfer masks, so favor whichever seems
more accurate in terms of labeling. I'm not sure what analysis you
will be doing, but our group has used this for structural
connectivity and resting-state functional connectivity analysis.
Since you are using longitudinal data, if you are doing any
connectivity analysis, you should look into minimizing ROI size
bias by transforming your T1s from each collection timepoint to an
average (see fMRIPrep for an easy implementation of this).