How accurate are machine learning models in predicting anti-seizure medication responses: A systematic review

Area of Science:

• Neurology
• Artificial Intelligence
• Pharmacogenomics

Background:

• Epilepsy management often involves sequential anti-seizure medication (ASM) trials, potentially delaying optimal treatment.
• Machine learning (ML) offers a potential solution for predicting individual patient responses to ASMs.
• Predicting ASM efficacy and tolerability is crucial for personalized epilepsy care.

Purpose of the Study:

• To review the effectiveness and limitations of ML models in predicting and classifying patient responses to ASMs.
• To assess how different data inputs impact the performance of ML prediction models for epilepsy.
• To synthesize current knowledge on ML applications in ASM response prediction.

Main Methods:

• A comprehensive literature search was conducted using PubMed and Scopus databases.
• Studies published up to November 2024 utilizing ML models for ASM response prediction were included.
• A systematic review methodology was employed to analyze the selected studies.

Main Results:

• 37 studies were included in the review, utilizing diverse data types like clinical information, MRI, EEG, and genetic data.
• Tree-based algorithms and Support Vector Machines were the most common ML models employed.
• Model performance varied significantly, from near-perfect accuracy to random chance, with data quality and quantity identified as key limitations.

Conclusions:

• ML models hold significant potential for predicting ASM responses in epilepsy, advancing precision medicine.
• Further research is essential to refine ML models for practical clinical application and enhance prediction accuracy.
• Addressing data limitations is critical for the successful implementation of ML in epilepsy treatment decisions.