Classification of Complete Myocardial Infarction Using Rule-Based Rough Set Method and Rough Set Explorer System

Abstract

In this study, a computerized diagnosis system is developed using Rough set classifier from multi-lead ECG signal for detection as well as the classification of five different types of myocardial infarction (MI) disease. The pathological features of ECG such as Inverted T-wave, ST segment deviation, or pathological Q wave, which are seen during MI, are extracted. An Information table and the knowledgebase are expanded from these pathological features after getting feedback from the cardiologist as well as consulting different medical books. The Information table contains 36 features and 341objects which include normal and five types of MI such as Anterior (AN), Inferior (IN), Antero lateral (ANLA), Inferior lateral (INLA), and Antero septal (ANSE) are used for assessment. The proposed system determines the degree of attributes dependency and their significance to find a smaller set of attributes, called reduct, alike the original set to predict the appropriate decision rules for MI classification. The robustness is justified by the “five-fold cross” validation technique using RSES tools. Finally, the proposed classifier illustrates its outperformance over the existing approaches in terms of sensitivity (99.75%), and accuracy (99.8%) for MI detection and 99.8% accuracy for MI classification.

Keywords:

• Accuracy
• ECG
• Knowledge bases
• MI
• Reduct
• RSES
• Rough set
• Sensitivity
• Specificity
• Train and test

Additional information

Funding

The part of the work is funded by the University Grant Commission (UGC), Govt. of India. The authors are thankful to UGC for providing financial assistance.

Notes on contributors

B. Halder

Basudev Halder received his BE (Computer Science & Engg) and ME (Computer Science & Engg) from University of Jadavpur, West Bengal, India. At present, he is a faculty member at Computer Sc & Engg department of Neotia Institute of Technology, Management and Science, Jhinga, West Bengal (India). He has been pursuing his PhD (Tech) from the University of Calcutta, India. His research interests in the area of biomedical signal processing which include embedding and retrieving patient’s information within ECG signal, feature extraction from ECG signal, classification and identification of heart disease from the ECG signal.E-mail: [email protected]

S. Mitra

Sucharita Mitra received her BSc Physics (H), MSc in electronic science and PhD (Tech) degree from the University of Calcutta. At present, she is assistant professor in the Department of Electronics of Netaji Nagar Day College. Formerly, she attached to the Indian Statistical Institute as a Post-Doctoral Fellow and Department of Applied Physics (CU) as a guest faculty. She visited the University of Oxford as Post-Doctoral fellow under the DBT-CREST award scheme of Govt. of India. Her research interests include biomedical signal and image processing, telemedicine, pattern recognition, etc., She published 55 research papers in reputed international and national journals and conference proceedings. Corresponding author. Email: [email protected]

M. Mitra

Madhuchhanda Mitra was born in Kolkata, India in 1961. She received her B Tech, MTech and PhD (Tech) degrees in 1987,1989 and 1998, respectively, from University of Calcutta, Kolkata, India. At present, she is professor in the Department of Applied Physics, University College of Technology, University of Calcutta, India, where she has been actively engaged in both teaching and research. She is the co-author of 150 research papers pertaining to the problems of fault analysis, biomedical signal processing, data acquisition and processing and material science. She is a recipient of “Griffith Memorial Award” of the University of Calcutta.E-mail: [email protected]