NEURAL NETWORK MODEL FOR DIAGNOSING MYOCARDIAL INFARCTION

Abstract

Body surface potential mapping (BSPM) is a non-invasive and effective method for diagnosing coronary heart disease (CHD) and acute myocardial infarction (AMI). However, most existing systems of BSPM are unable to create standard diagnostic criteria. Aim. To develop the neural network model (NNM) for diagnosing Q-wave AMI and to assess the model effectiveness. Material and methods. The BSPM method in 90 leads was used in 96 controls, 35 patients with anterior Q-wave AMI, 43 with posterior Q-wave AMI, 14 with inferior Q-wave AMI, and 21 with lateral Q-wave AMI. The input NNM layer was decomposed into five subsets corresponding to horizontal levels of registered signals, using amplitudes of Q, R, S, and T waves and the ST segment. The output layer produced the probability of the norm (controls) and different AMI locations. Results. Exploring the NNM performance in controls and AMI patients, sensitivity of 100% and specificity of 97,4% was observed. Sensitivity reached 100% for anterior Q-wave AMI, 94,4% for posterior Q-wave AMI, 85,7% for inferior Q-wave AMI, and 83,3% for lateral Q-wave AMI. Conclusion. Our data have demonstrated the effectiveness of NNM in AMI diagnostics.

Keywords

• body surface potential mapping
• neural network modeling
• acute myocardial infarction