A novel vectorcardiogram tensor decomposition method achieved 99.80% total accuracy and >0.88 AUC for localizing 11 types of myocardial infarction.
Does an automated method based on Tucker decomposition accurately localize myocardial infarction using vectorcardiograms?
A novel tensor decomposition-based algorithm using vectorcardiograms achieved 99.80% accuracy in localizing 11 types of myocardial infarction.
OBJECTIVE: Myocardial infarction (MI) causes rapid and permanent damage to the heart muscle. Therefore, it can deteriorate the myocardial structure and function if not timely diagnosed and treated. However, it is difficult to determine the precise localization of MI based on vectorcardiogram (VCG) due to the existing studies ignore the spatiotemporal features of VCG. METHODS: In this paper, a precise MI localization method was proposed based on Tucker decomposition. The multi-scale characteristics of wavelet transform and the spatiotemporal characteristics of VCG were used to construct the VCG tensor containing the local and the spatiotemporal information. The VCG tensor was compressed in the time dimension based on Tucker decomposition to remove redundant information and extract the local spatiotemporal features. The features were fed back to the TreeBagger classifier. RESULTS: The proposed method achieved a total accuracy of 99.80% for 11 types of MI on the benchmark Physikalisch-Technische Bundesanstalt database. The area under the receiver operating characteristic curves and precision-recall curves of each kind of VCG signal was more than 0.88. CONCLUSION: The proposed algorithm effectively realized the classification of normal and 11 categories of MI using VCG. SIGNIFICANCE: Therefore, this study provides new ideas for the intelligent diagnosis of MI based on VCG.
Zhang et al. (Tue,) conducted a other in Myocardial infarction. Tucker decomposition-based MI localization method was evaluated on Total accuracy for 11 types of MI. A novel vectorcardiogram tensor decomposition method achieved 99.80% total accuracy and >0.88 AUC for localizing 11 types of myocardial infarction.