Causal Brain Network Predicts Surgical Outcomes in Patients With Drug-Resistant Epilepsy: A Retrospective Comparative...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Epilepsy Research

Background:

• Surgical success rates for drug-resistant epilepsy (DRE) remain variable (30%-70%).
• Predicting surgical outcomes is crucial for optimizing treatment strategies.
• Network neuroscience offers potential biomarkers for surgical prognostication.

Purpose of the Study:

• To systematically explore causal brain network biomarkers for predicting DRE surgical outcomes.
• To validate the efficacy of convergent cross mapping (CCM) in identifying predictive network features.
• To develop machine learning models for outcome prediction based on identified biomarkers.

Main Methods:

• Retrospective analysis of electrocorticogram (ECoG) data from 17 DRE patients.
• Construction of causal brain networks using six algorithms on ictal ECoG within the epileptogenic zone (EZ) and non-epileptogenic zone (NEZ).
• Application of Mann-Whitney-U-test and machine learning (SVM) for biomarker identification and outcome prediction.

Main Results:

• Alpha-frequency band (8-13 Hz) causal connectivity in the EZ, calculated by CCM, showed significant differences between successful and failed surgery groups (P=7.85e-08, Cohen's d=0.77).
• CCM-defined EZ brain networks effectively distinguished surgical outcomes, even when considering clinical covariates.
• A support vector machine (SVM) classifier achieved an average prediction accuracy of 84.48% using these network features.

Conclusions:

• Causal brain network analysis, particularly CCM-derived alpha-band connectivity in the EZ, serves as a reliable biomarker for predicting DRE surgical outcomes.
• Machine learning models integrating these network features can accurately forecast surgical success.
• These findings can guide clinical decision-making and improve patient management in DRE.