Analysis of Functional MRI Signals by Using Approximate Spectral Clustering based on a Geodesic Measure of Similarity

Abstract

Resting State-fMRI represents an emerging and powerful tool to explore brain functional connectivity (FC) changes associated with neurologic disorders. Compared to activation/task-related fMRI, Resting State fMRI has the advantages such as (i) Blood oxygen level dependent (BOLD) fMRI signals are self-generated and independent of subject's performance during the task and (ii) a single dataset is sufficient to extract a set of resting state networks that allows to explore whole brain FC. In the experiment, 200 dynamic 3D-tomograms were taken from healthy 11 male and 9 female volunteers (aged 20-41 years) using T2 * weighted echo-planar imaging method with the Philips Achieve 1.5 T MRI system SENSE-head 8 coil and TE=40 ms, deviation angle 90 with a temporal resolution of 2.64 seconds and a spatial resolution of 3.5x3.5x4 mm This study will help to determine the various functional neurologal connections of the human brain in the resting state and use approximate spectral clustering based on a geodesic measure of similarity to define different neurological disordes created by these differences in functional neurological connections.