Can we predict anti-seizure medication response in focal epilepsy using machine learning?

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Focal epilepsy treatment response to anti-seizure medications (ASMs) varies significantly.
• Predicting ASM response is crucial for optimizing patient care and outcomes.

Purpose of the Study:

• To assess the feasibility of a machine learning (ML) model using clinical factors and diffusion tensor imaging (DTI) to predict ASM response in focal epilepsy.
• To compare the predictive power of clinical factors versus DTI-derived metrics.

Main Methods:

• A retrospective analysis of 160 newly diagnosed focal epilepsy patients who underwent DTI.
• Support vector machine (SVM) algorithm with k-fold cross-validation was employed.
• Analysis included clinical characteristics, conventional DTI measurements, and structural connectomic profiles.

Main Results:

• An SVM model using clinical factors achieved 87.5% accuracy and an AUC of 0.882.
• DTI measures (conventional and connectomic) showed lower predictive accuracy (62.5% and 68.7%, respectively).
• Clinical factors demonstrated significantly higher predictive performance compared to DTI metrics.

Conclusions:

• Machine learning, particularly using clinical factors, is a viable approach for predicting ASM response in focal epilepsy.
• Clinical information is a more critical predictor of ASM response than DTI-based neuroimaging in this patient cohort.