The idea is to apply spangy directly onto a sphere where the curvature information has been mapped. There are tons of methods to map a mesh onto a sphere but ywe could use the one in Lefevre & Auzias, 2015 that can be implemented rapidly in slam
https://hal.archives-ouvertes.fr/hal-01222933/file/74-Lefevre.pdf
It is not easy to compare the numerical eigenvectors with the analytic ones because the eigenvalues are not unique (the eigenspace is degenerated).
But we could use the inverse mapping to back-project the spherical harmonics onto the brain mesh and check whether
a. the functions are (almost) orthogonal and
b. they allow to obtain comparable results when looking at the spectral bands.
The idea is to apply spangy directly onto a sphere where the curvature information has been mapped. There are tons of methods to map a mesh onto a sphere but ywe could use the one in Lefevre & Auzias, 2015 that can be implemented rapidly in slam https://hal.archives-ouvertes.fr/hal-01222933/file/74-Lefevre.pdf