Open adamltyson opened 9 months ago
Tagging @niksirbi to see if you have any specific recommendations for human atlases we should support.
Probably the best place to start would be various versions of the "MNI" (Montreal Neurological Institute) templates. They come bundled with all major (f)MRI analysis software and are very widely used.
The confusing bit is that there are many different "MNI" templates, and they also come in multiple versions. This webpage provides a good explainer of various "MNI" spaces.
I would probably start with the 2009 versions of the ICBM 152 nonlinear average templates.
If we managed to write a script to fetch them via TemplateFlow (which is what I would also recommend) and convert them to BrainGlobe Atlas API format, we could batch-import the lot of them.
However, there is a catch. In human MRI, it is very common to mix and match templates and annotations, so there is no such thing as an MNI "atlas" in the BrainGlobe sense (template + annotations). We would need to decide how to combine multiple templates with multiple annotations. In other words, this may force us to implement some of the existing ideas for BrainGlobe Atlas API 2.0.
Could we leverage TemplateFlow somehow?