An object oriented approach to multimodal imaging data in neuroscience

We propose a methodological framework for exploring complex multimodal imaging data from a neuroscience study with the aim of identifying a data-driven group structure in the patients sample, possibly connected with the presence/absence of lifetime mental disorder. The functional covariances of fMRI signals are first considered as data objects. Appropriate clustering procedures and low dimensional representations are proposed. For inference, a Frechet estimator of both the covariance operator itself and the average covariance operator is used. A permutation procedure to test the equality of the covariance operators between two groups is also considered. We finally propose a method to incorporate spatial dependencies between different brain regions, merging the information from both the Structural Networks and the Dynamic functional activity.