Disclosing brain functional connections with multi-dimensional approaches

In the last decades, it has become clear that functional interactions between brain regions are crucial for understanding brain's functioning. These interactions can be assessed through functional connectivity approaches that characterize statistical dependencies between time series obtained by time-resolved neuroimaging techniques. Nevertheless, these techniques provide information about brain activity at the scale of brain regions with an extent in the order of several millimetres that constitute a functional unit and are constituted by multiple voxels. To apply conventional bivariate connectivity metrics, the signal of voxels in a brain region (a multi-dimensional set) is usually reduced to a single signal per brain region with a potential loss of information. To overcome this limitation, we have developed a family of “multi-dimensional connectivity” methods that take direct advantage of the multidimensional nature of the data. Here, we present a collection of these methods and show, in simulations, that they outperform the approach based on dimensionality reduction and non multidimensional connectivity methods.