Figure 1.

Overall workflow. The first step involves collecting ECG tests, manual CAC scores and clinical data from eligible patients. The ECG features are then automatically extracted using the commercially available GE MUSE software. Top ECG features were selected using the LASSO feature selection method and used for constructing eRiS. M_nom was constructed using clinical features and eRiS. M_ecg and M_nom were validated for prognostic performance and predicting downstream MACE events. LASSO: least absolute shrinkage and selection operator; eRiS: ECG risk score, Mecg: Cox PH model using eRiS alone; Mnom: Nomogram with eRiS, CAC and clinical factors.