Classification of cardiac disorders using 1D local ternary patterns based on pulse plethysmograph signals

Heart diseases are a major cause of human casualties each year. An accurate and efficient diagnosis is essential to minimize their risk. This paper presents a system for the classification of multiple cardiac disorders based on pulse plethysmographic (PuPG) signal analysis. In particular, the work focuses on the detection and classification of ischemic and rheumatic heart diseases using proposed 1D local ternary patterns of PuPG signals. The proposed methodology is applied on a self-collected dataset consisting of 250 PuPG signals from 50 normal/healthy subjects, 140 signals from 28 ischemic patients, and 180 signals from 36 rheumatic patients. The effectiveness of proposed feature descriptors to precisely represent different classes of data is verified by several classifiers namely K-nearest neighbours (KNN), support vector machines (SVM), and decision tree (DT) with 10-fold cross-validation. The proposed methodology achieves the best detection performance with 99% accuracy, 100% sensitivity, and 98% specificity using an SVM classifier with cubic kernel. The comparative analysis demonstrates that the proposed method is more accurate and reliable as compared to several existing works on cardiac disorders classification.