Closed danjgale closed 5 years ago
Turns out adding probabilistic atlases is not as simple as using NiftiMapsMasker
. 4D probabilistic atlases from nilearn add several extra parameters and some API changes that complicate, rather than simplify, the process and code. Because 4D probabilistic atlases are so user-specific, niimasker will focus on labelled deterministic atlases (3D). Therefore, probabilistic atlases must be first thresholded and converted to a 3D image with appropriate labels.
Right now niimasker extracts data via
NiftiMasker
orNiftiLabelsMasker
, which assumes that each voxel has a single region assignment (value). nilearn has several probabilistic atlases which can be used with NiftiMapsMasker. So, this enhancement is dependent on #1, and would be a great follow-up to that upon completion.