Automatic classification of seizure adequacy in electroconvulsive therapy: A proof-of-concept study

Machine learning trained on raw EEG from 1,000 ECT sessions in 130 patients classifies seizure adequacy into null-response (0), threshold (T), and adequate (A) with AUC >0.9 for 0 vs. T and T vs. A.

• Why it matters: ECT dosing depends on manual seizure-quality assessment, which is subjective and varies across patients, limiting objective, personalized decision support.
• What they did: The study labeled EEGs by expert review into actionable categories (0, T, A), trained multiple algorithms on raw EEG, and evaluated classification accuracy using ROC-based AUC.
• The result: The best model discriminated 0 vs. T and T vs. A with AUC >0.9, but predicting seizure quality in subsequent treatments was only moderately accurate, and adding clinical variables did not improve performance.

AI summary of the abstract · may contain errors · not medical advice · read the paper.