Automated stenosis classification on invasive coronary angiography using modified dual cross pattern with iterative feature selection

Coronary artery disease (CAD) is a global health concern; the need for early diagnosis cannot be overstated. Many machine learning techniques have been used electrocardiography (ECG) signal to detect CAD and they have used advanced signal processing methods. In this study, we present an automated novel approach for detecting coronary artery stenosis, by integrating the residual exemplar center symmetric dual cross pattern (ResExCSDCP), relief and iterative neighborhood component analysis (RFINCA) techniques. In this work, we collected three coronary angiography images datasets to show general classification ability of the proposed model and these images were gathered from right coronary artery (RCA), left anterior descending artery (LAD), and circumflex artery (CX). The features have been extracted from patches by deploying ResExCSDCP feature extractor. The most informative features have been selected deploying RFINCA and k-nearest neighbor (kNN) has been employed for classification. Our proposed ResExCSDCP and RFINCA-based model attained accuracies of 96.73% ± 1.38, 97.24% ± 1.12%, and 98.51% ± 0.31% for the automatic detection of RCA, LAD, and CX coronary angiography images, respectively. The results demonstrate that our proposal has the potential to assist the cardiologists in making accurate diagnosis and improve the quality of cardiac health.