Methodological Issues in Evaluating Machine Learning Models for EEG Seizure Prediction: Good Cross-Validation Accuracy Does Not Guarantee Generalization to New Patients

There is an increasing interest in applying artificial intelligence techniques to forecast epileptic seizures. In particular, machine learning algorithms could extract nonlinear statistical regularities from electroencephalographic (EEG) time series that can anticipate abnormal brain activity. The recent literature reports promising results seizure detection and prediction tasks using deep methods. However, performance evaluation often based on questionable randomized cross-validation schemes, which introduce correlated signals (e.g., EEG data recorded the same patient during nearby periods of day) into partitioning training test sets. present study demonstrates use more stringent strategies, such as those leave-one-patient-out partitioning, leads a drop accuracy about 80% 50% for standard eXtreme Gradient Boosting (XGBoost) classifier two different Our findings suggest definition rigorous protocols crucial ensure generalizability predictive models before proceeding clinical trials.