[ENH] Surface GLM - Support multiple runs in a GLM

[smeisler]

Is there an existing issue for this?

• I have searched the existing issues

Describe your proposed enhancement in detail.

Related to neurostars post: https://neurostars.org/t/glm-on-surface-data-in-nilearn/22632

It would be good if Nilearn could run GLMs on surface data with multiple BOLD runs. Currently, the only way to run a GLM on surface data (nilearn.glm.first_level.run_glm), which only accepts one BOLD run and design matrix per instance.

Benefits to the change

More flexible processing for multi-run surface analyses.

Pseudocode for the new behavior, if applicable

I imagine an easiest, but not most stylish, change would be to have run_glm accept an iterable object for [X] and [Y]. If the length of these iterable objects are 1, then run the function as normal. If not, apply whatever fixed effect models are used for the multi-run GLMs on BOLD data.

Thoughts?

[bthirion]

Thx for opening this. What we need is a SurfaceFirstLevelModel object that can has an API similar to the FirstLevelModel. We'll do that as soon as we have a decent SurfaceMasker.

[Remi-Gau]

Will close this one as completed by #4126.

See examples:

• https://github.com/nilearn/nilearn/blob/main/examples/08_experimental/plot_localizer_new_surface_analysis_experimental.py
• https://github.com/nilearn/nilearn/blob/main/examples/08_experimental/plot_new_surface_bids_analysis_experimental.py

Feedbacks are welcome.