Using Spectral and Temporal Filters with EEG Signal to Predict the Temporal Lobe Epilepsy Outcome after Antiseizure Medication via Machine Learning

Abstract

Abstract Epilepsy is a neurological disorder in which transient alteration of brain. Predicting outcomes in epilepsy is essential since the prediction could provide feedback that can foster improvement in the outcomes. This study aimed to investigate whether applying spectral and temporal filters to resting-state electroencephalogram (EEG) signals could improve the prediction of patients' outcomes after antiseizure medication for temporal lobe epilepsy (TLE). We collected EEG data from a total of 46 patients (seizure-free (SF, n = 22) or nonseizure-free (NSF, n = 24)) with TLE and reviewed their clinical data retrospectively. We dissected spectral and temporal ranges with various time-domain features (Hjorth parameters, statistical parameters, energy, and zero-crossing rate) and compared their performance by applying optimal frequency only, optimal duration only, and both. For all time-domain features, optimal frequency and time strategy (OFTS) showed the highest performance in distinguishing SF patients from NSF patients (0.759 ± 0.148 AUC). In addition, the best performance using statistical parameters as a feature vector was a frequency band of 39–41 Hz at a window length of 210s, with an AUC of 0.748. By identifying the optimal parameters, we improved the prediction model’s performance. These parameters can function as standard parameters for outcome prediction using resting-state EEG signals.