Visualization of surface maps
mwcole opened this issue · 2 comments
I found an issue in /tools/map_to_surface.py, which is used to map results to the Glasser surface parcels. The problem is that the surface visualization is slightly different between different versions of the Glasser parcels. I'm not sure why they differ slightly, but it's important to have the border visualization match the parcels. I just fixed it by switching map_to_surface.py to use this file: ColeAnticevicNetPartition/CortexSubcortex_ColeAnticevic_NetPartition_wSubcorGSR_parcels_LR.dlabel.nii
(It was using Q1-Q6_RelatedParcellation210.LR.CorticalAreas_dil_Colors.32k_fs_RL.dlabel.nii)
@ito-takuya, since you originally created this function, can you verify that this was an appropriate change? The data lined up with the CAB-NP borders when I made this change.
Short answer: yes. But in reality, it depends. In theory, the atlas used to remap back to surface should always be the atlas that was used to parcellate the data to begin with. For the purposes of this toolbox (and the parcellation function provided with this toolbox) this is the correct change.
But, for example, the HCP352 data set that I had preprocessed was originally parcellated and preprocessed using the Q1-Q6 parcellation, which has flipped hemispheres (R->L, rather than L->R). And so any surface mapping using that data set needs to use the original Glasser Q1-Q6 atlas when mapping back to a surface visualization. (The reason the borders are off is that the mappings are mapped to the wrong hemisphere.) I think an equivalent alternative would be to swap from first 180 parcels with the last 180 parcels, and then use the current function. An important corollary is that any analysis that uses the HCP352 data set needs to swap the network definitions from cab-np (since those definitions assume L->R).