MultiviewICA is a novel independent component analysis model for group studies. The likelihood of our model is available in closed form. We develop an alternate quasi-newton method to optimize it. We show that it recovers meaningful common sources on fMRI and MEG data. On fMRI data, our model demonstrates improved sensitivity in identifying common sources among subjects. On MEG data, it yields more accurate source localization. Applied on 200 subjects (Cam-CAN data) it reveals a clear sequence of evoked activity in sensor and source space.
Code: https://github.com/hugorichard/multiviewica
Paper: https://arxiv.org/abs/2006.06635
Youtube Presentation (10 minutes): https://www.youtube.com/watch?v=aBzrUc7hHyo
Recommended citation: Richard, H., Gresele, L., Hyvärinen, A., Thirion, B., Gramfort, A., & Ablin, P. (2020). Modeling Shared Responses in Neuroimaging Studies through MultiView ICA. arXiv preprint arXiv:2006.06635. https://arxiv.org/pdf/2006.06635.pdf
