Functional imaging methods such as resting-state fMRI allow to describe interactions among different areas of the brain, thus deriving a functional connectivity matrix of the entire brain network. Tracking functional relationships among different regions of interest can be applied, besides a pure modelling perspective, also to discovering procedures to detect brain diseases and anomalies, or pursuing rehabilitation of subjects with structural damages. However, network characterization is often regarded as frequency-independent, so that the frequency at which interactions take place among different regions is ignored. In this paper, we show how simple filtering procedures over different bands, applied to the resting-state fMRI signals, result in highly different connectivity matrices. Thus, it is highlighted that the functional network can be significantly dependent on the considered frequency range for the fMRI signal. This both justifies the need for a careful filtering of the signals, that avoids filtering out relevant frequencies, and also hints the possibility of classifying functional interactions according to the frequency where the connectivity among two areas is the strongest.
