Detection of myocardial infarction in 12 lead ECG using support vector machine

• Published in: Applied soft computing Vol. 64; pp. 138 - 147
• Main Authors: Dohare, Ashok Kumar, Kumar, Vinod, Kumar, Ritesh
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.03.2018
• Subjects: Composite lead; Electrocardiogram; Myocardial infarction; Principal component analysis (PCA); Statistical parameters support vector machine; Myocardial infarction; Statistical parameters support vector machine; Composite lead; Electrocardiogram; Principal component analysis (PCA)
• ISSN: 1568-4946; 1872-9681
• DOI: 10.1016/j.asoc.2017.12.001

Detection of Myocardial Infarction in 12 lead ECG using Support Vector Machine. [Display omitted] •Detection of myocardial infarction (MI) method is proposed based on SVM classifier using 12-lead ECG system.•Two hundred twenty parameters determined in average beat of selecting 10 s data length data in all 12-lead ECG of the standard PTB database.•The MI detection sensitivity is 96.66%, specificity 100% and accuracy 98.33% with original features 220.•Applying PCA reduction method features from 220 to 14 features achieved MI detection with SVM classifier is: sensitivity 96.66%, specificity 96.66% and accuracy 96.66%.•The proposed method for MI detection perform results of MI detection is comparable and higher. In this paper, we propose myocardial infarction (MI) detection using 12-lead ECG data and analysis of each lead with the help of composite lead. This composite lead is used to detect ECG wave components and clinical wave intervals in all the 12-lead ECG. The four clinical features such as P duration, QRS duration, ST-T complex interval and QT interval are globally determined from average beats of all the 12-lead ECG. Then peak to peak amplitude, area, mean, standard deviation, skewness and kurtosis are determined for P duration, QRS duration and ST-T complex interval of average beats of all the 12-lead ECG. These 220 (4 + 6 × 3 × 12) parameters are used for myocardial infarction detection. The standard 12-lead ECG data of 60 myocardial infarction subjects and 60 healthy controls (HC) cases are obtained from Physikalisch-Technische Bundesanstalt (PTB) database and tested with support vector machine (SVM) classifier. The MI detection sensitivity, specificity and accuracy are 96.66%, 100% and 98.33% respectively. To reduce the computational complexity, feature dimension reduction is important. Therefore, proposed method applies Principal Component Analysis (PCA) reduction technique. In this proposed method, 220 features are reduced to 14 features, using these 14 features, MI detection achieved by SVM classifier is: sensitivity 96.66%, specificity 96.66% and accuracy 96.66%.