Abstract

Functional connectivity estimation between brain regions is a challenging problem. Indeed, both noise and the possible heterogeneity of the brain regions lead to either under- or overestimation of the inter-regional correlation. While some existing methods handle either of these issues, none tackle both at the same time. Popular methods include correlating regional averages, which is not robust to low average intra-regional correlation, and averaging pairwise inter-regional correlations, which is not robust to high noise. Here I will discuss our novel clustering-based non-parametric estimator of the inter-regional correlation, which simultaneously offsets the impact of noise and arbitrary intra-regional correlation through an intermediate aggregation. I will also describe both empirical validation on synthetic data and illustrations on real-world rat and human rs-fMRI datasets that further demonstrate its effectiveness. By design, our method provides as well an empirical distribution of the inter-correlation. I will then present an ongoing work introducing the concept of distribution-weighted connectivity networks and applying it on real-world data.