A Deep-Learning Neural Network-Based Predictive System for the Occurrence of Major Adverse Cardiovascular Events (MACE) in Patients with Acute Myocardial Infarction

Abstract

Deep-learning is an emerging technology in health informatics nowadays. Therefore, this paper proposes a novel Deep Neural Network (DNN)-based diagnosis system for Cardiovascular Disease (CVD) in patients with Acute Myocardial Infarction (AMI). In this research, Korea Acute Myocardial Infarction Registry (KAMIR-IV) dataset is used and 11,189 subjects are extracted after data pre-processing, and then divided into two subdatasets such as males’ and females’ datasets. Later, all datasets are splitted into training and test datasets, and consequently, the Synthetic Minority Oversampling Technique (SMOTE) on training data for data imbalance problem has been applied. The proposed prediction model is trained on oversampled training data, and hyperparameters are tuned using grid search approach. Following, the performance of proposed model is evaluated using performance measures such as accuracy, precision, recall, F1-score, and the Area under the ROC Curve (AUC). The proposed DNN-based prediction model achieved an accuracy of 0.9835, a precision of 0.9835, a recall of 0.9835, an F1-score of 0.9834, and an AUC of 0.9943 on a complete dataset whereas, the accuracy of 0.9713, a precision of 0.9710, a recall of 0.9713, an F1-score of 0.9710, and an AUC of 0.9989 on males’ subdata and an accuracy of 0.9607, a precision of 0.9701, a recall of 0.9613, an F1-score of 0.9720, and an AUC of 0.9985 on females’ subdata. In addition, a web-based decision support system is developed and deployed on the local server for physicians, doctors, and CVD patients. Consequently, our finding was that the proposed diagnosis system is predicting efficiently for all patients and diagnosing the major adverse cardiovascular events’ (MACE) occurrences accurately in order to select the proper treatment for patients with AMI.