A Simple and Effective Method for Detecting Myocardial Infarction Based on Deep Convolutional Neural Network

Myocardial infarction (MI) is the main cause of sudden death in patients with cardiovascular diseases (CVD), thus timely detection of myocardial infarction is crucial for saving patients' lives. This paper presents an algorithm based on deep convolution neural network (CNN) to detect myocardial infarction, using electrocardiogram (ECG) signal from lead II. The algorithm proposed in this paper uses neither manual feature extraction nor feature selection, and instead of performing heartbeat segmentation, the method takes 3 second ECG signal segments as input. For our experiments, we conduct two datasets of denoised ECG set and original ECG set to corroborate the robustness of the algorithm to noise in ECG signal. We evaluate the model by a 10-fold cross-validation on the PTB database and achieve the state-of-the-art result: accuracy = 99.34%, sensitivity = 99.79% and specificity = 97.44% for the denoised ECG signal, and accuracy = 98.59%, sensitivity = 99.53% and specificity = 94.50% for the raw ECG signal.