template asymmetry

[tknapen]

Hi devs,

awesome tool this. We've immediately moved towards using it as standard. One thing we noticed is that there is a seeming asymmetry in the template cuts, such that the left hemisphere calcarine cut runs deeper than the one on the right hemisphere. This results in the right hemisphere having a nice rectangular V1 patch based on polar angle maps (couple of examples shown), while the left hemisphere template cuts result in a sometimes discontinuous and less rectangularly shaped V1. The severity of this effect varies across subjects (see sub-01 in the polar maps for a rather extreme case), but I believe the most likely cause is the template cuts.

This effect also shows up as a displacement of the V123 fovea in eccentricity maps (shown below). There is a purple band along the cuts in the eccentricity maps for the RH, whereas in the left hemisphere the black foveal region is closer to the cut - a sign that the calcarine cut encroaches on the V1 fovea by 3-5mm too much.

Sorry to drop this on you like this, but since the template is encoded in a json file and isn't super-easily altered (by me at present, at least) perhaps you have tools to edit the json for the template calcarine cuts in the left hemisphere?

Best,

Tomas Knapen

[mvdoc]

Hi @tknapen, sorry but for some reason I didn't get a notification and this issue flew under my radar.

It is fairly easy to change the reference template cuts. They are stored in json format just to cache them, but they can be edited on fsaverage with the usual tools. Currently I am just using the cuts we have on fsaverage provided in pycortex. If you have different cuts on fsaverage, you can create a new template following this script https://github.com/gallantlab/autoflatten/blob/main/scripts/make_fsaverage_cut_template.py Then you can pass the new template file with the --template-file option, see https://gallantlab.org/autoflatten/detailed-options/#custom-templates

The tricky part here though is whether the projection of the cuts from fsaverage to individual subjects (via freesurfer's surface2surface anatomical mapping) would actually match the subject-specific retinotopy maps. I don't think the result would be as precise as creating cuts based on the actual retinotopy localizer. But maybe slightly editing the left hemisphere would be sufficient? I'd be happy to add any new improvement to the template.