Cortical and Network Reorganization in Glioma-Related Epilepsy: Insights From Structural and Machine Learning Analyses

Abstract

Background: Epilepsy is a common symptom in patients with diffuse lower-grade glioma (DLGG). However, the specific role of cortical alterations in glioma-related epilepsy (GRE) remains unclear. This study is aimed at investigating the reorganization of cortical architecture and network changes associated with GRE.

Materials and Methods: High-resolution T1-weighted and T2-weighted images were acquired from patients with DLGG (GRE = 68, non-GRE = 79) and 94 healthy controls (HCs). Cortical thickness and myelin content were calculated using the Human Connectome Project pipeline. Characteristics of structural covariance networks were computed using graph theory and network-based statistic. Cortical thickness, myelin content, and network characteristics were compared among three groups. A GRE individual prediction model was constructed using an automated machine learning approach.

Results: Compared with HCs, both GRE and non-GRE groups exhibited cortical thinning in the tumor ipsilateral hemisphere, whereas there was cortical thickening in the contralateral hemisphere. Regarding the connectome characteristics, both GRE and non-GRE groups showed decreased nodal efficiency and connections in multiple regions. When comparing GRE with non-GRE, the GRE group exhibited more pronounced cortical thickening and demyelination in the contralateral orbitofrontal gyrus and superior frontal gyrus, with further decreased connections in the sensorimotor network, default mode network, and salience network. Finally, an XGBoost model based on cortical features enabled classification of GRE individuals with an accuracy of 0.80 and an AUC of 0.87.

Conclusion: These findings deepen our understanding of the comprehensive cortical alterations in patients with DLGG and simultaneously provide novel insights into the potential pathophysiological mechanisms underlying GRE.