Cardiovascular Disease Prediction Using Modified Version of ResNet-50 Model

Cardiovascular disease (CD) is responsible for the biggest number of deaths globally when compared to other illnesses. The early and accurate diagnosis of diseases can help save many patients’ lives. Cardiovascular disease can be detected using medical tests such as an electrocardiogram (ECG) signal, heart sounds, and computed tomography (CT) images. Recognizing CD using ECG signals is essential compared to all the above tests. For the early detection of cardiac patients, many researchers analyzed the cardiovascular disease risk factors and employed machine learning or deep learning approaches. However, these models still need to be enhanced. In this study, we compared two models of CNN techniques which are LeNet-5 and ResNet-50 then the highest-performing model was selected. We proposed a modified version of the ResNet-50 model to improve the prediction of cardiovascular disease patients. The ECG data from the patients were taken into account as the crucial input for the CD classifier. Evaluation metrics such as accuracy, precision, recall, f1-score, and specificity are used to compare the modified version to the original version. The obtained results showed that the performance of the ResNet-50 model was better than the LeNet-5 model. Modifying the ResNet-50 model’s architecture improved the model’s ability to analyze the ECGs dataset and extract its most significant features. According to the results, the modified ResNet-50 model based on the ECG image dataset increased the accuracy compared to the original ResNet-50 model by 4.83 percentage points.