Use of an artificial intelligence-enabled electrocardiogram for screening people without coronary heart disease
Objective:To explore the application of artificial intelligence-assisted electrocardiograms (ECG) to identify patients without coronary heart disease (coronary artery stenosis < 50%).Methods:Patients with suspected coronary heart disease and who underwent coronary angiography during hospitalization were enrolled. The ECG data set was established based on standard 12-lead ECG. The ECG was labeled as a group without coronary disease and a control group according to whether the main coronary artery or its main branches were narrowed by less than 50% in diameter. A deep neural network model was established by training ECG to identify patients without coronary heart disease.Results:A total of 4 489 ECG medical records were included, of which 4 187 were used for model construction and 302 were used for external verification of the model. The area under curve (AUC) value, sensitivity, and specificity of the model were 0.70, 0.701, 0.630, and 0.469 respectively. The AUC value of external validation was 0.55, sensitivity 0.359, specificity 0.784, and F1 score 0.373.Conclusion:The artificial intelligence model based on ECG can identify patients without coronary heart disease in patients with suspected coronary heart disease, which has certain clinical application value.
万方数据
