Deep learned representations of the resting 12-lead electrocardiogram to predict V̇O2 at peak exercise.

AIM: To leverage deep learning on the resting 12-lead ECG to estimate peak oxygen consumption (V̇O2  PEAK) without cardiopulmonary exercise testing (CPET).METHODS: V̇O2  PEAK estimation models were developed in 1,891 individuals undergoing CPET at Massachusetts General Hospital (age 45±19 years, 38% female) and validated in a separate test set (MGH Test, n=448) and external sample (BWH Test, n=1,076). Three penalized linear models were compared: a) age, sex, and body mass index ("Basic"), b) Basic plus standard ECG measurements ("Basic + ECG Parameters"), and c) Basic plus 320 deep learning-derived ECG variables instead of ECG measurements ("Deep ECG-V̇O2"). Associations between estimated V̇O2  PEAK and incident disease were assessed using proportional hazards models within 84,718 primary care patients without CPET.RESULTS: Inference ECGs preceded CPET by 7 days (median, interquartile range 27-0 days). Among models, Deep ECG-V̇O2 was most accurate in MGH Test (r=0.845, 95%CI 0.817-0.870; mean absolute error [MAE] 5.84, 95%CI 5.39-6.29) and BWH Test (r=0.552, 95%CI 0.509-0.592, MAE 6.49, 95%CI 6.21-6.67). Deep ECG-V̇O2 also outperformed the Wasserman, Jones, and FRIEND reference equations (p<0.01 for comparisons of correlation). Performance was higher in BWH Test when individuals with heart failure were excluded (r=0.628, 95%CI 0.567-0.682; MAE 5.97, 95%CI 5.57-6.37). Deep ECG-V̇O2 estimated V̇O2  PEAK <14 mL/kg/min was associated with increased risks of incident atrial fibrillation (hazard ratio 1.36 [95%CI 1.21-1.54]), myocardial infarction (1.21 [1.02-1.45]), heart failure (1.67 [1.49-1.88]), and death (1.84 [1.68-2.03]).CONCLUSIONS: Deep learning-enabled analysis of the resting 12-lead ECG can estimate exercise capacity (V̇O2  PEAK) at scale to enable efficient cardiovascular risk stratification.