Our paper did in fact project it to the volume, by taking into account the non-overlap between subjects, and the individual nature of the original parcellations, and the result is a label volume that is generally thinner than typical cortex, and has holes "missing" where there was no cortical area that was more likely than white matter or CSF. See supplemental figure S4 (which has a link to the data files).
If you want to use the fine-grained detail of our cortical parcellation, we strongly recommend that you use processing methods that preserve a similar level of detail and tissue accuracy, and for group analysis that generally means surface-based methods. Note that for a single individual, you can in fact go back and forth between volume and surface without appreciable losses in localization (only resampling losses, effectively). The problem with group-average volume cortical data is the fact that different subjects aren't all that well aligned (even simple cortical overlap generally isn't great), and averaging those voxels that are only somewhat aligned is where you lose your localization. There are also some low-variability parts of cortex like the insula and CeS that volume alignment does reasonably with, and the subcortical gray matter structures (which are not in the HCP MMP 1.0) are also well-aligned in the volume. Tim On Tue, Apr 2, 2019 at 1:32 PM Mor Regev <more...@gmail.com> wrote: > Thanks Tim! I understand that, but can't it be back projected? > > Mor > > On Tue, Apr 2, 2019 at 2:23 PM Timothy Coalson <tsc...@mst.edu> wrote: > >> The HCP MMP 1.0 parcellation could not have been made without using >> surface-based methods, due to their increased accuracy in aligning >> functional areas over existing volume-based registrations. Volume-based >> group data generally cannot have the cortical precision that the HCP MMP >> 1.0 implies. See our paper showing this: >> >> https://www.ncbi.nlm.nih.gov/pubmed/29925602 >> >> Tim >> >> >> On Tue, Apr 2, 2019 at 11:41 AM Mor Regev <more...@gmail.com> wrote: >> >>> Hello, >>> >>> I would like to use Glasser's parcellation in volume space. Is there an >>> available nifti I could use? >>> >>> Thanks, >>> Mor >>> >>> >>> -- >>> Dr. Mor Regev >>> Montreal Neurological Institute >>> McGill University >>> 3801 University St >>> <https://maps.google.com/?q=3801+University+St+Montreal,+QC+Canada+H3A2B4&entry=gmail&source=g> >>> Montreal, QC Canada H3A2B4 >>> <https://maps.google.com/?q=3801+University+St+Montreal,+QC+Canada+H3A2B4&entry=gmail&source=g> >>> >>> _______________________________________________ >>> HCP-Users mailing list >>> HCP-Users@humanconnectome.org >>> http://lists.humanconnectome.org/mailman/listinfo/hcp-users >>> >> _______________________________________________ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users
