Supervised classification of structural brain networks reveals gender differences

The human brain anatomical organization can be modeled as a complex network, called a connectome. This representation enables the use of powerful mathematical tools to detect individual differences in brain structure. In this paper, we propose three methods based on Support Vector Machines, to classify connectomes according to a given label. We demonstrate the efficiency of our methods compared to existing approaches through experiments on a dataset built from publicly available MRI data, and for the classification between male and female brains. We further interpret the classification outcome in regards to one particular wiring feature: the inter-hemispheric ratio.