Multistage Electrocardiographic Profiling for Mental Stress-Induced Myocardial Ischemia in Women: A Prospective, Age-Matched Cohort Study in China

BACKGROUND: Mental stress-induced myocardial ischemia is often clinically silent and associated with increased cardiovascular risk, particularly in women. Conventional ECG-based detection is limited, while the reference standard, positron emission tomography/computed tomography, is costly and less accessible. We aimed to characterize mental stress-induced myocardial ischemia using multistage ECG analysis and machine learning to improve early, noninvasive detection. METHODS: We enrolled 2 female cohorts (18-75 years) from a single-center tertiary hospital in Guangzhou, China: a study cohort (2020-2021) and an independent test cohort (2022-2023). Participants included women with angina, nonobstructive coronary artery disease, and age-matched healthy controls. Multistage ECGs were recorded during 3 mental stress tasks and segmented into Rest, Stress, and Recovery phases. A total of 88 interpretable ECG variables, including morphological indices and heart rate variability, were extracted from each stage, and interstage differences were then calculated. Mental stress-induced myocardial ischemia labeling was obtained from positron emission tomography/computed tomography. Machine learning classifiers (K-nearest neighbors, logistic regression, random forest, support vector machine, extreme gradient boosting) were trained on the study cohort and evaluated in the test cohort using single-stage, multistage, and interstage features. RESULTS: