Hybrid Deep Learning algorithm for Abnormal Heart Beat detection using ECG Signals

According to the World Health Organization (WHO), heart disease is increasing worldwide due to various factors such as lifestyle, food habits, and lack of physical activity. Recently, even adults and children have been affected by heart disease. The abnormal heartbeat analysis and classification are crucial for proper treatment. The electrocardiogram (ECG) is globally used for analyzing heart function. However, the manual inspection of ECG signals is very difficult and time-consuming. Many research works focus on ECG signal analysis to detect and classify abnormal heartbeats. ECG signals are unstable and vary from person to person based on age and other factors, making effective feature extraction challenging. Moreover, achieving the required accuracy remains a challenge. To address this issue, a novel hybrid Deep Learning (DL) model called Transformer with Multihead Attention (MHA)-Bidirectional Long Short Term Memory (BiLSTM) is proposed. The Transformer captures long-range dependencies, while the BiLSTM extracts sequential dependencies from the ECG signal. The integration of these two models not only extracts important features but also maintains the temporal structure, leading to improved classification accuracy. The proposed DL model is compared with traditional standalone models such as Transformer, BiLSTM, and Convolutional Neural Network (CNN). For evaluation, the MIT-BIH dataset is used, which consists of normal and four types of abnormal heartbeats. The experimental results show that the proposed network effectively classifies abnormal heartbeats with an accuracy of 98.93%. In comparison, the Transformer model achieves 96.27%, CNN 93.73%, and BiLSTM 95.47%. Additionally, the proposed network is compared with recent research works. From the comparison and evaluation, it is evident that the proposed network is suitable for highly reliable abnormal heartbeat detection. This can be helpful for doctors in detecting cardiac diseases in minimal time.