An Optimization Approach to EEG Feature Extraction for the Prediction of Neurological Outcome

As part of the George B. Moody PhysioNet Challenge 2023, our team, Swarthbeat, developed a computational approach that uses electroencephalograms (EEGs) to predict the neurological recovery of patients following cardiac arrest. Our method involved selecting a small number of significant features from a much larger set by optimization based on the PhysioNet Challenge score. Significance was determined using a bagged tree ensemble method. The model for our highest ranking entry was trained using this smaller set of features with adaptive boosting. Our model received a Challenge score of 0.52 (17th out of 36 ranked teams) on the hidden test set.