Developing Deep Learning Model for Abnormal Heartbeat Detection in Electrocardiography Signals

The electrocardiogram (ECG) is an essential instrument used by physicians that provides information about cardiovascular health and is the gold standard for detecting abnormal heartbeats. Traditional medical tests have limitations, which include delayed diagnoses and a high potential for misdiagnosis, especially when after the attention to cardiovascular health grows. Given the importance of early detection, medication, and rehabilitation for cardiovascular diseases (CVD), the rate of misinterpreting must be reduced. This research study develops a CNN-BLSTM (Convolutional Neural Network and Bidirectional Long Short-Term Memory), hybrid Deep Learning (DL) model for categorizing both normal and abnormal cardiac data. The proposed technique is evaluated using ECG signals from the MIT-BIH arrhythmia database, which comprises both normal and five types of abnormal signals. Further, this research study compares the performance of our model to that of existing transfer learning models such as ResNet-50 and AlexNet. With an overall accuracy of 98.11%, the proposed CNN-BLSTM model surpasses both ResNet-50 and AlexNet. Furthermore, the model delivers the greatest prediction scores for each type of heartbeat classification, indicating its effectiveness in accurate and comprehensive heartbeat signal analysis.