Myocardial Infarction Detection Based on Multi-lead Ensemble Neural Network

• Published in: Conference proceedings (IEEE Engineering in Medicine and Biology Society. Conf.) Vol. 2019; pp. 2614 - 2617
• Main Authors: Wang, H.M., Zhao, W., Jia, D.Y., Hu, J., Li, Z.Q., Yan, C., You, T.Y.
• Format: Conference Proceeding Journal Article
• Language: English
• Published: United States IEEE 01.07.2019
• Subjects: Algorithms; Convolution; Diagnosis, Computer-Assisted; Electrocardiography; Feature extraction; Humans; Kernel; Myocardial Infarction - diagnosis; Myocardium; Neural networks; Neural Networks, Computer; Sensitivity and Specificity
• ISSN: 1557-170X; 1558-4615
• DOI: 10.1109/EMBC.2019.8856392

Automatic myocardial infarction (MI) detection using an electrocardiogram (ECG) is of great significance for improving the survival rate of patients. In this study, we propose a multi-lead ensemble neural network (MENN) to distinguish anterior myocardial infarction (AMI) and inferior myocardial infarction (IMI) from healthy control (HC) respectively. In the study, three kinds of sub-networks and multi-lead ECG signals are combined, which fully explores the information of ECG signals and improves the classification performance. The algorithm is evaluated on the PTB database by 5-fold inter-subject cross-validation and the sensitivity (Se), specificity (Sp) and area under the curve (AUC) of AMI detection are 98.35%, 97.49%, 97.92%; The Se, Sp, and AUC of IMI detection are 93.17%, 92.02%, 92.60%. The proposed method achieves the state of the art results on both tasks and outperforms the baseline methods. Hence, the proposed method is potential for automatic MI diagnosis.