BISeizuRe: BERT-Inspired Seizure Data Representation to Improve Epilepsy Monitoring

TL;DR

This paper introduces BENDR, a BERT-inspired EEG model trained on large datasets and fine-tuned for seizure detection, significantly reducing false positives and improving generalization in epilepsy monitoring.

Contribution

It presents a novel BERT-based EEG model with a two-phase training process, including extensive pre-training and subject-specific fine-tuning, to enhance seizure detection accuracy.

Findings

Achieved as low as 0.23 false positives per hour

Model outperforms baseline by 2.5 times in reducing FP/h

Effective pre-training and fine-tuning strategies improve generalization

Abstract

This study presents a novel approach for EEG-based seizure detection leveraging a BERT-based model. The model, BENDR, undergoes a two-phase training process. Initially, it is pre-trained on the extensive Temple University Hospital EEG Corpus (TUEG), a 1.5 TB dataset comprising over 10,000 subjects, to extract common EEG data patterns. Subsequently, the model is fine-tuned on the CHB-MIT Scalp EEG Database, consisting of 664 EEG recordings from 24 pediatric patients, of which 198 contain seizure events. Key contributions include optimizing fine-tuning on the CHB-MIT dataset, where the impact of model architecture, pre-processing, and post-processing techniques are thoroughly examined to enhance sensitivity and reduce false positives per hour (FP/h). We also explored custom training strategies to ascertain the most effective setup. The model undergoes a novel second pre-training phase before subject-specific fine-tuning, enhancing its generalization capabilities. The optimized model demonstrates substantial performance enhancements, achieving as low as 0.23 FP/h, 2.5$\times$ lower than the baseline model, with a lower but still acceptable sensitivity rate, showcasing the effectiveness of applying a BERT-based approach on EEG-based seizure detection.