Automated detection of myocardial infarction from ECG signal using variational mode decomposition based analysis

Abstract

In this Letter, the authors propose a variational mode decomposition method for quantifying diagnostic information of myocardial infarction (MI) from the electrocardiogram (ECG) signal. The multiscale mode energy and principal component (PC) of multiscale covariance matrices are used as features. The mode energies determine the strength of the mode, and the PCs provide the representation of the ECG signal with less redundancy. K-nearest neighbour and support vector machine classifier are utilised to assess the performance of the extracted features for the detection and classification of MI and normal (healthy control). The proposed method achieved a specificity of 99.88%, sensitivity of 99.90%, and accuracy of 99.88%. Experimental results demonstrate that the proposed method with the multiscale mode energy and PC features achieved better output compared to the previously published work.

Keywords

• feature extraction
• signal classification
• principal component analysis
• medical signal processing
• covariance matrices
• medical signal detection
• support vector machines
• electrocardiography
• support vector machine classifier
• classification
• mi
• multiscale mode energy
• pc
• automated detection
• myocardial infarction
• ecg signal
• variational mode decomposition method
• diagnostic information
• electrocardiogram signal
• principal component
• multiscale covariance matrices
• mode energies
• neighbour