Diagnostic information extractin from automated ECG analysis in anterior myocardial infarction using neural networks

摘要

The diagnostic algorithm in currently available automated ECG analysis mainly uses the "if-then" rule, according to which it is diffiult to extract ambiguous medical information from ECG data sufficiently enough to make an accurate diagnosis. There is a large number of patients who are falsely diagnosed a having cardiac diseases by automated ECG analysis, despite of the fact that they do not actually suffer from the diseases. The authors have devised a new differentiation technique by applying neural networks to ECG diagnosis of anterior myocardial infarctin,which is high in both sensitivity and specificity. in this study, we evaluated the performance of this neural network system in differentiating cases with anterior myocardial infarction from those without the disease, using patients who had been classified into the category of "abnormal" by automated ECG analysis and were later found on reexamination by cardiologists to be included in the category of "not abnormal." The system correctly differentiated between the two categories with a recognition rate of 91.9