Brain Structural Covariance Networks of Patients with Migraine: A Source-based Morphometry Study

Abstract

Migraine is associated with grey matter changes in the human brain, but the interrelationships between voxels and naturally grouped patterns of structural variations need to be identi ed to understand how migraine a ects the brain at the network level. Three dimensional, T1 weighted images of 45 migraine patients and 46 healthy controls who underwent brain MR scanning in a 3 Tesla scanner were selected. The data were analyzed using Computational Anatomy Toolbox-CAT12. Group independent component analysis (Group ICA) was performed on grey matter volumes using GIFT toolbox. Two sample t-tests were performed using component loadings of each component to nd out the signi cant independent components (ICs). Nine maximally independent components (ICs) were resulted using group ICA and labeled according to the Resting State Network (RSN) atlas. Among them 6 ICs were found to be signi cant (p<0.05) based on two-sample ttests representing the attention, sensorimotor, frontal, and visual networks. Further, mask based multivariate pattern analyses (MVPA) were performed to distinguish patients with migraine and healthy controls. MVPA revealed that component 4 (sensorimotor, classi cation accuracy=61.53%), component 5 (sensorimotor, classi cation accuracy=73.62%), component 7 (sensorimotor, classi cation accuracy=69.23%) and component 8 (visual, classi cation accuracy=68.13%) show di erent potentials to correctly classify patients with migraine and healthy subjects (p<0.05, number of permutations, n=1000). Source Based Morphometry can detect the structural covariance networks of brains with migraine, which exhibit signi cant di erences when compared with healthy controls. Further, the above structural network changes can be used to develop an e ective biomarker for objective diagnosis of migraine.