Advantages of using Empirical Mode Decomposition and Hilbert Transformation for Delineating Resting State Functional Brain Networks

The goal of the resting-state functional connectivity studies is to determine the inherent dynamics of the brain networks while the body is at rest. These networks get differentially activated when the brain is involved in various tasks such as processing of sensory inputs, initiating motor activities, or various cognitive tasks. Resting state functional connectivity networks are commonly revealed by determining Pearson Correlation Coefficients of the Blood Oxygenation Level Dependent (BOLD) signals collected from different brain regions using functional Magnetic Resonance Imaging (fMRI) while the subject is not actively performing any task. However, the functional connectivity thus determined does not correlate well with the known structural connectivity between different brain regions. Here, we used Empirical Mode decomposition (EMD), followed by Hilbert Transformation (HT), to determine the resting state functional connectivity in the human brains. We show the advantage of using this EMD-HT method using somatomotor network as an example. We show that the time series data decomposed by this method improves correlation of the derived functional connectivity with the known structural connectivity (especially for low -TR fMRI data) as compared to the methods commonly used.