Minimum spanning tree analysis for epilepsy magnetoencephalography (MEG) data

Abstract

Aim: Recently, brain network research is actively conducted through the application of graph theory. However, comparison between brain networks is subject to bias issues due to topological characteristics and heterogeneity across subjects. The minimum spanning tree (MST) is a method that is increasingly applied to overcome the thresholding problem. In this study, the aim is to use the MST analysis in comparing epilepsy patients and controls to find the differences between groups. Methods: The MST combines entities for epileptic magnetoencephalography (MEG) data. The MST was applied and compared to 21 left surgery (LT) and 21 right surgery (RT) patients with epilepsy and good postoperative prognosis and a healthy control (HC) group. MST metrics such as betweenness centrality, eccentricity, diameter, and leaf fraction, are computed and compared to describe the integration and efficiency of the network. The MST analysis is applied to each subject, and then the integrated MST is obtained using the distance concept. This approach can be advantageous when comparing the topological structure of patients to controls with the same number of nodes. Results: The HC group showed less topological change and more network efficiency than the epilepsy LT and RT groups. In addition, the posterior cingulate gyrus was found as a hub node only in the patient group in individual and integrated subject data analysis. Conclusions: This study suggests propose that the hippocampus borrows from the default network when one side fails, compensating for the weakened function.