Figure 1

Summary of the proposed method. We start (top-left ovals) from data-driven structural (functional) connectomes, derived from dMRI (fMRI) scans. Each functional connectome evidences the correlation between the neural activity of different brain areas, and then the presence of approximately synchronized clusters, formed by areas with high correlation. There is a degree of uncertainty on the number of clusters the areas can be best aggregated into. Each structural connectome reflects the brain anatomy, with some uncertainty in the weight of each connection. Right (green) ovals: model-based method used to bridge functional and structural connectivity. The reference (homogeneous) network model has an initial topology defined by the structural connectome and node dynamics imposed by a NMM. The uncertainty on the clusters is managed through a multi-resolution approach, by analyzing different cluster partitions. For each resolution level, the uncertainty on the connection weights is reduced by optimizing them to enforce the existence of the cluster synchronous solution of that level. Bottom-left oval: one resolution level is chosen, and the corresponding model is made heterogeneous, thus obtaining again approximate synchronization.