Computer-Aided Diagnosis System for Early Prediction of Atherosclerosis using Machine Learning and K-fold cross-validation

Abstract

Atherosclerosis known as coronary artery disease (CAD) becomes epidemic in any society that relies on an industrial-technological system with an associated behavioral alteration in people's lifestyles as junk food consumerism and stressful habits. However, this disease residue the first cause of death in industrialized countries, despite many new therapeutic approaches and risk factors prevention. Moreover, atherosclerosis misdiagnosis has side costly effects. In this paper, we have proposed a computer-aided diagnosis system based on K-Nearest Neighbors (KNN) and Artificial Neural Network (ANN) algorithms. Then, we applied K-fold cross-validation in order to split the databases and reach the best model with the higher accuracy and fewer side effects. In this proposed work, we tested the reached model on 573 patients with several effective features which collecting from Cleveland and Z-Alizadeh Sani datasets. Then Area Under the Curve (AUC), F1-Score, and accuracy were used to enrich and determine the effectiveness of each predictive model. Using Machine Learning (ML) methods, K-fold cross-validation, and performance evaluation metrics, 96.78% average accuracy is achieved with the original training accuracy of 100%, which means the prediction system is obtained as the best predictive model comparing to the previous studies.