Efficient Myocardial Infarction Detection Using Hybrid CNN-BiLSTM with Depthwise Separable Convolution on ECG Signals

Myocardial infarction is a cardiovascular disease caused by blocked blood vessels that restrict blood flow to the heart, potentially leading to cardiac death. Early detection and accurate classification are essential for timely medical intervention. Although hybrid deep learning methods show promise in detecting myocardial infarction using electrocardiogram signals, challenges remain regarding data imbalance, limited preprocessing, and inefficient model structures. To address these issues, this study proposes the Efficient Myocardial Infarction Detection model, which combines a convolutional neural network and bidirectional long short-term memory with hyperparameter optimization, depthwise separable convolution, and advanced preprocessing using a low-pass Butterworth filter and discrete wavelet transform. Additionally, we develop comparison models using convolutional recurrent and convolutional long short-term memory architectures. Experimental results show that the proposed model outperforms the others, achieving high performance in both training and testing datasets. In the training set, the model reaches accuracy, precision, sensitivity, specificity, and F1-score of 94.54 percent, 94.13 percent, 94.10 percent, 99.40 percent, and 94.54 percent, respectively. On the testing set, the scores are 94.41 percent, 94.00 percent, 93.99 percent, 99.38 percent, and 93.96 percent. These results demonstrate the model's robustness and potential for clinical application.