Hybrid BiGRU-BiLSTM Model for Real-Time ECG Arrhythmia Detection Using Wearable Sensors

Abstract

Accurate and real-time detection of cardiac arrhythmias is essential for timely medical intervention. Advances in wearable devices and deep learning have made it feasible to continuously monitor electrocardiogram (ECG) signals, facilitating early identification of abnormal heart rhythms. For arrhythmia detection, this study presents a hybrid deep learning architecture combining Bidirectional Gated Recurrent Units (Bi-GRU) and Bidirectional Long Short-Term Memory (Bi-LSTM). To improve the extraction and classification of relevant features, the model incorporates Dilated Convolutional Neural Networks (DCNNs) alongside a hierarchical attention mechanism. The proposed framework achieves a maximum accuracy of 99.97%, surpassing the performance of conventional approaches. The proposed model achieved an accuracy of 99.97%, with a precision of 99.91%, a recall of 99.88%, and an F1-score of 99.89%. The hierarchical attention mechanism enhances interpretability by highlighting significant ECG segments contributing to classification decisions, ensuring transparency in clinical analysis. This method is well-suited for real-time implementation in wearable cardiac monitoring systems.