Prediction of Long-Term Postsurgical Seizure Recurrence From MRI Brain Hub Disruption in Patients With Temporal Lobe...

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Temporal lobe epilepsy (TLE) often leads to long-term seizure recurrence despite initial surgical success.
• Disruption of highly connected brain regions (hubs) may impact surgical outcomes in TLE.
• Predicting long-term seizure outcomes is crucial for patient management.

Purpose of the Study:

• To investigate whether disruption of normative brain hubs predicts long-term seizure recurrence in TLE patients.
• To develop and validate machine learning models for predicting seizure outcomes using network disruption biomarkers.
• To assess the clinical utility of network-level biomarkers for postoperative risk stratification.

Main Methods:

• Prospective, multimodal cohort study of 175 drug-resistant TLE patients undergoing surgery.
• Derivation of structural and functional connectomes using MRI and fMRI.
• Quantification of patient-specific hub disruption using the participation coefficient measure.
• Training and validation of machine learning models incorporating network measures, clinical, and demographic data.

Main Results:

• A multimodal approach combining structural and functional connectomes outperformed unimodal and clinical-only models.
• The model achieved high specificity (80.0%) and moderate-to-high negative predictive value (63.9%) in predicting seizure recurrence.
• Disruption in the hippocampi and dorsal attention network hubs predicted long-term seizure recurrence.

Conclusions:

• Disruption of normative hub architecture provides interpretable biomarkers for TLE seizure outcomes.
• The developed model shows potential for clinical utility in postoperative risk stratification and patient counseling.
• Network-level biomarkers can complement conventional predictors for improved TLE management.