EEG-Based Inter-Patient Epileptic Seizure Detection Combining Domain Adversarial Training with CNN-BiLSTM Network

TL;DR

This paper tackles cross-patient epileptic seizure detection from EEG by addressing inter-patient domain shift. It introduces a two-stage CNN-BiLSTM framework where stage 1 applies domain adversarial training to the CNN to learn patient-invariant features, and stage 2 integrates a BiLSTM to model temporal seizure evolution, followed by end-to-end fine-tuning. Evaluated on EEG data from 20 focal epilepsy patients with leave-one-patient-out validation, the approach shows superior performance over non-adversarial baselines and effectively leverages temporal dependencies to reduce false positives. While promising, the method encounters challenges with EEG patterns that are highly atypical (e.g., in infancy), suggesting avenues for richer datasets and further optimization of temporal windows to enhance clinical applicability.

Abstract

Automated epileptic seizure detection from electroencephalogram (EEG) remains challenging due to significant individual differences in EEG patterns across patients. While existing studies achieve high accuracy with patient-specific approaches, they face difficulties in generalizing to new patients. To address this, we propose a detection framework combining domain adversarial training with a convolutional neural network (CNN) and a bidirectional long short-term memory (BiLSTM). First, the CNN extracts local patient-invariant features through domain adversarial training, which optimizes seizure detection accuracy while minimizing patient-specific characteristics. Then, the BiLSTM captures temporal dependencies in the extracted features to model seizure evolution patterns. Evaluation using EEG recordings from 20 patients with focal epilepsy demonstrated superior performance over non-adversarial methods, achieving high detection accuracy across different patients. The integration of adversarial training with temporal modeling enables robust cross-patient seizure detection.

Figures (4)

• Figure 1: Schematic overview of the proposed CNN-BiLSTM framework with adversarial training
• Figure 2: Averaged loss progression during the first-stage training across patients as a function of training epochs. (a) Label loss for seizure prediction. (b) Domain loss for patient classification.
• Figure 3: Preprocessed EEG signals and seizure detection results from CNN-BiLSTM models with adversarial training (w/ AT) and without adversarial training (w/o AT). (a) Patient K. (b) Patient D. Gray regions indicate seizure periods annotated by an epileptologist. Horizontal dashed lines in the seizure prediction probability panels represent the detection threshold.
• Figure 4: Patient-specific MCC comparison between with and without adversarial training (AT)