Application of ensemble classifiers in accurate diagnosis of myocardial ischemia conditions

Desai, Usha and Nayak, Gurudas C and Seshikala, G (2017) Application of ensemble classifiers in accurate diagnosis of myocardial ischemia conditions. Progress in Artificial Intelligence. ISSN 2192-6352

[img] PDF
2864.pdf - Published Version
Restricted to Registered users only

Download (1MB) | Request a copy


Myocardial ischemia (MI) is abnormal cardiac condition during which myocardium is deprived of blood and oxygen supply. It is the most prominent heart disease around the globe, and its prevalence increases with age factors. Accurate and timely identification and treatment of MI help to prevent from series of dreadful ischemic heart diseases such as heart attack and heart failure. However, the manual interpretation of electrocardiogram (ECG) for diagnosis of MI is tedious task for physicians. Hence, automated classification is very useful for diagnosis of MI condition. Our study presents an approach for automated diagnosis of MI condition using ECG beats. In this methodology, non-linearity in ECG is finely captured using higher-order statistics (HOS), and then, dimensionality is reduced. Confidence interval of features is verified using Student’s t-test (probability (p)-value <0.05). Decision Tree (DT) as the base classifier following different ensemble classifiers, viz. Adaboost (ADB), Bagging (BGG), Rotation Forest (ROF) and Random Forest (RAF) are subjected for diagnosis of normal sinus rhythm (NSR) against MI conditions. Among these, ROF is diagnosed with 99.51 ± 0.31% accuracy, 98.82±1.04% sensitivity and 99.83±0.15% specificity. In this system, higher specificity rate defines precision in diagnosis of True Negative (TN) condition. Results achieved in B Usha Desai 1 Department of Electronics and Communication Engineering, Nitte Mahalinga Adyanthaya Memorial Institute of Technology, Nitte Karkala, Udupi, Karnataka 574110, India 2 School of Electronics and Communication Engineering, REVA University, Bengaluru 560064, India 3 Department of Instrumentation and Control Engineering, Manipal Institute of Technology, Manipal University, Manipal 576104, India this study are efficient with 0.9886 of Matthews correlation coefficient (MCC). Our proposed system is non-invasive and cost-efficient and helps in screening of MI, predominantly in rural and remote areas.

Item Type: Article
Uncontrolled Keywords: Computer-aided-diagnosis · Coronary artery disease · European ST-T database · ECG preprocessing · Ensemble classifiers · Matthews correlation coefficient
Subjects: Engineering > MIT Manipal > Instrumentation and Control
Depositing User: MIT Library
Date Deposited: 07 Jul 2017 06:07
Last Modified: 07 Jul 2017 06:07

Actions (login required)

View Item View Item