Multi support vector machine and image processing for diagnosis of coronary artery disease

Abstract

The optimal non-invasive test, Coronary computed tomography angiography (CCTA), is to control coronary artery disease (CAD). This paper proposes a developed algorithm called Multi Support Vector Machine (MSVM) applied in classification and diagnosing a common heart disease, CAD, utilizing the features extracted from the patients’ CCTA images through two image-processing-based approaches. These image-processing-based approaches including the quantification of cardiovascular vessels and the autoencoder (AE) network are utilized for the extraction of the features from the CCTA images. Then, a novel MSVM algorithm is developed for diagnosing heart diseases. A dataset from the Tehran Heart Center is utilized in addition to a collection of datasets from the literature to evaluate the performance of the proposed algorithms based on accuracy, precision, and recall performance measures. The proposed MSVM algorithm is compared with a number of existing methods in the literature where the results show that the proposed MSVM algorithm outperforms all the competing methods in terms of all the performance measures. In addition, it is concluded that the proposed MSVM algorithm performs much better than the classical SVM method under all the scenarios.