GSP Analysis of Brain Imaging Data from Athletes with History of Multiple Concussions

Abstract

Study of neurological disorders affecting the structure-function relationships in the brain has been an ongoing challenge in neuroscience. Joint analysis of structure and function of the brain may disentangle a number of mechanisms and operations that can help interpret the interdependence between white matter degeneration and degradation of cognitive abilities. In this scenario, graph signal processing analysis of different signals generated within the physical structure of the brain may provide new insights and corroborate existing clinical findings. This paper illustrates the utility of graph signal processing tools in the joint analysis of diffusion and functional magnetic resonance imaging (i.e. dMRI and fMRI) data collected from a population of former athletes with a history of multiple concussions, and healthy subjects. Specifically, the distributions of the energy of low-graph-frequency components of the functional networks (derived from fMRI) are observed to be significantly different for fronto-temporal regions of the brain in athletes and healthy subjects. Furthermore, for the two groups of subjects, we observe significantly different associations between the ages of subjects and the energies of high graph frequency components in lingual region. While the effect on fronto-temporal regions for former athletes is in line with the existing clinical studies on concussion, significantly different associations between age and features extracted using GSP for the two groups of subjects could inform future clinical applications and medical diagnosis.