Parameterized Aperiodic and Periodic Components of Single-channel EEG Enable Reliable Seizure Detection

Abstract

Abstract Objective: While it is clinically important, a reliable and economical solution to automatic seizure detection for patients at home is yet to be developed. Traditional algorithms rely on multi-channel EEG signals and features of canonical EEG power decomposition. This study is aimed to parameterize the power spectra of EEG signals about their aperiodic and periodic components, and to examine the effectiveness of these novel features of a single-channel EEG for seizure detection. Methods: We employed the publicly available multi-channel CHB-MIT Scalp EEG Database to gauge the effectiveness of our approach. We first adopted a power spectra parameterization method to characterize the aperiodic and periodic components of the ictal and inter-ictal EEGs and systematically performed the statistical analysis on parameters of these two characteristic components, by channel and by patient. We then tested the effectiveness of four highly discriminative features for automatic seizure detection using a support vector machine on a single-channel EEG selected for each patient. The performance of our algorithm was compared to those systems of comparable complexity (using one or two channels of EEG), in terms of accuracy, specificity, sensitivity, precision, and F1 score. Results: Some channels of EEG for each patient show strikingly different distributions of the offset and exponent parameters characterizing the aperiodic components between the ictal and inter-ictal EEGs. Similarly, the two highest power of the periodic components (PW1 and PW2) also show significant differences. The total power (TPW1 and TPW2) at the frequencies corresponding to PW1 and PW2 demonstrate even greater statistical significance between the ictal and inter-ictal states. The seizure detection algorithm based on four features (offset, exponent, TPW1, and TPW2) offers a sensitivity of 97.7%, specificity of 99.5%, accuracy of 99.4%, precision of 97.5%, and F1 score of 97.4%. Significance: A new approach to epileptic EEG feature extraction can better characterize the ictal and inter-ictal EEG signals and result in efficient and effective seizure detection based on a single channel of EEG.