Epileptic Seizures Detection with Low Processing Load Using the R-R interval of ECG Signals and Deep Neural Network

Early detection of seizure occurrence is important in improving the lives of patients with epilepsy. Although various methods have been presented so far, there are still challenges regarding the ease of using these methods. This paper aims to present an approach based on neural networks to detect epileptic seizures using ECG signals instead of EEG signals. In the proposed method, only single-lead ECG signals are used for detecting seizures, making them highly suitable for wearable instruments. In the proposed approach, light pre-processing is first performed to prepare raw ECG signals as labeled data. Then, R-R intervals are extracted and considered as features for classification. Without using other manual feature extraction methods, a combined structure of a convolutional neural network and a bidirectional long short-term memory network is employed for automatic feature extraction and classification. The proposed method was evaluated using a strict Leave-One-Subject-Out cross-validation on five patients, achieving a mean accuracy of ۸۲.۴%, sensitivity of ۷۴.۴%, specificity of ۹۱.۳%, and an aggregated AUC of ۰.۸۴۴. The model requires approximately ۴۵۰۰ trainable parameters, confirming its low computational load. This research demonstrates that a single-lead ECG signal, combined with a lightweight deep learning algorithm, can be effectively used in monitoring systems for epilepsy patients.

کلیدواژه ها:

Epileptic seizure ، Detection ، deep neural network ، ECG signals ، R-R interval