Study of the inter-individual variability of intrinsic connectivity data: detection of unstable networks and sub-populations in a three-dimensional table.

We propose two methodologies to better understand the inter-individual variability of functional Magnetic Resonance Imaging brain data. The aim is to quantify whether the "average" dendrogram is representative of the initial population and to identify its possible sources of instability. The first method identifies networks that can lead to unstable partitions of the "average" dendrogram. The second approach identifies homogeneous sub-populations of subjects for whom their associated "average" dendrograms are more stable than that of the original population. These two methods will be illustrated on simulated data from intrinsic connectivity data obtained by functional MRI. The two suggested approaches to detect an unstable network or the presence of sub-populations have shown good numerical behavior when the noise level does not mask the structure of the data.