Mesial temporal lobe epilepsy (MTLE) is the most common medically refractory focal epilepsy, classified into seizure-onset patterns (SOPs) such as hypersynchronous (HYP) and low-voltage fast (LVF) rhythms. However, the morphological and network differences underlying these SOPs and their prognostic implications remain unclear.Using voxel-based morphometry (VBM) and individualized structural covariance networks (IDSCNs), we analyzed 55 MTLE patients and 37 healthy controls with high-resolution MRI and stereotactic EEG recordings. VBM identified grey matter volume (GMV) abnormalities, and IDSCNs evaluated patient-specific network disruptions. Hierarchical clustering explored network subtypes, and ROC analysis assessed the predictive value of VBM and IDSCN measures for surgical outcomes.HYP-onset patients exhibited ipsilateral mesial temporal atrophy and widespread bilateral network disruptions centered on the hippocampus and supramarginal gyrus, along with significant interhemispheric connectivity abnormalities. LVF-onset patients showed subcortical abnormalities and hemispheric network disruptions centered on the fusiform gyrus, with limited interhemispheric propagation. Both SOPs displayed basal ganglia abnormalities. Hierarchical clustering revealed SOP-specific network subtypes. IDSCN outperformed VBM in predicting surgical outcomes.This study identifies distinct SOP-related morphological and network abnormalities in MTLE and suggests that IDSCN may offer valuable insights into personalized treatment planning and prognostic evaluation, supporting further exploration of network-based biomarkers.Copyright © 2025 The Authors. Published by Elsevier Inc. All rights reserved.