A Two-Stage Automatic System for Detection of Interictal Epileptiform Discharges from Scalp Electroencephalograms.

Researchers

Journal

Modalities

Models

Abstract

To develop a deep learning-based automatic system with reliable performance in detecting interictal epileptiform discharges (IEDs) from scalp electroencephalograms (EEGs).For the present study, 484 raw scalp EEG recordings were included, standardized, and split into 406 for training and 78 for testing. Two neurophysiologists individually annotated the recordings for training in channel-wise manner. Annotations were divided into segments, on which 9 deep neural networks (DNNs) were trained for the multi-classification of IED, artifact and background. The fitted IED detectors were then evaluated on 78 EEG recordings with IED events fully annotated by 3 experts independently (majority agreement). A two-montage-based decision mechanism (TMDM) was designed to determine whether an IED event occurred at a single time instant. Area under the precision-recall curve (AUPRC), as well as false positive rates, F1 scores, and Kappa agreement scores for sensitivity = 0.8 were estimated.In multi-type classification, 5 DNNs provided one-vs.-rest AUPRC mean value >0.993 using 5-fold cross-validation. In IED detection, the system that had integrated the temporal convolutional network (TCN)-based IED detector and the TMDM rule achieved an AUPRC of 0.811. False positive was 0.194 per minute (11.64 per hour) and F1 score was 0.745. Agreement score between the system and the experts was 0.905.The proposed framework provides a TCN-based IED detector and a novel two-montage-based determining mechanism that combined to make an automatic IED detection system.The system would be useful in aiding clinic EEG interpretation.Significance StatementThis work has presented a deep learning-based system with a false positive of 0.194 per minute (11.64 per hour) for sensitivity = 0.8 on 78 whole clinical EEG recordings. Especially, these recordings were selected to challenge the system, we therefore would expect better performance in more general diagnostic scenario.We collected a sizable multi-institute dataset, and the 78 whole clinical EEG recordings for testing have been fully annotated by experts. We believe disclosure of this dataset would benefit researches in this field.Additionally, the DNNs were trained for the multi-classification of IED, artifact and background waveforms. By using this procedure, we attempt to not only improving performance, but also making a step forward to the ultimate automatic EEG interpretation.Copyright © 2023 Wang et al.