Association Between Predicted Cardiac Age and Right Ventricular Dysfunction Based on Electrocardiogram Artificial Intelligence in Acute Pulmonary Embolism

Abstract BACKGROUND: Deep neural network (DNN) has been used to estimate age from 12-leads electrocardiograms (ECGs). And the difference between electrocardiographic heart age (ECG-age) and chronological age was approved to be associated with adverse outcomes. We aimed to evaluate the prediction value of the discrepancies between ECG-age and chronological age (Δage) on right ventricular dysfunction (RVD) in patients with pulmonary embolism (PE). METHODS: The ECG-age prediction model, built on RegNet and trained using mean squared error (MSE), was developed with 21,799 ECG recordings from 18,869 patients in mimic dataset and had been validated on external test sets previously. In this study, the model was applied to the biological age among patients with PE diagnosed and hospitalized in China-Japan Friendship Hospital between 2016-2022. The difference between the ECG-age and biological age (Δage) of each patient was calculated. Correlation and logistic analyses were conducted to investigate the associations between Δage and echocardiographic parameters on RVD and pulmonary circulation. RESULTS: Contemporaneous (intervals of examinations dates less than 7 days) ECG signal and echocardiography data comprises the ECG-Echo data pairs. 86 data pairs of 74 patients were included. The median biological age was 69 (57,77) years and median ECG-age 57 (49,60) years. The median Δage was -9.5±12.2 years. Δage was significantly associated with tricuspid regurgitation velocity (R = 0.38, P = 0.01), systolic pulmonary arterial pressure (R = 0.35, P = 0.035), main pulmonary artery diameter (R = 0.34, P = 0.005) and thickness of interventricular septum (R= 0.25, P = 0.027). (Fig.1) The prevalence of RVD in Δage ≥0 group and <0 group were 6(25%) and 9(18.75%), respectively(P=0.76). Logistic regression revealed Δage ≥ 0 and biological aging were risk factors for RVD [adjusted ORs 8.39 (1.31,53.58) and 1.08(1.02,1.14), respectively]. CONCLUSIONS: The increased difference between ECG-age and biological age was associated with RVD and other parameters of pulmonary circulation. Given the wide availability and low cost of ECG, ECG-age promises to be a new biomarker of risk classification and prognosis in PE patients with the further validation.