Abstract
Coronary microvascular disease (CMD), particularly prevalent among women, is associated with increased morbidity and mortality, making clinical screening vital for effective management. However, limited publicly available screening-level data hinders disease-specific biomarker discovery. To address this gap, 80 female angina patients without obstructive coronary artery disease and 40 age-matched female controls were prospectively enrolled to curate a new dataset. All participants underwent adenosine stress with electrocardiogram (ECG) monitoring across Rest, Stress, and Recovery stages. CMD diagnosis was confirmed with the standard clinical criterion, i.e., coronary flow reserve (CFR) < 2.0 via PET/CT. Using ECG variables from different stages, we developed machine learning models to classify CMD, thus validating dataset's effectiveness in CMD identification. We also validated the potential of ECG for differential diagnosis through joint analysis with the published mental stress-induced myocardial ischemia (MSIMI) dataset, which is based on the same cohort under different stress conditions. Disease-specific ECG variable sets were identified. Our findings highlight the value of multi-stage ECG in CMD screening. We expect this dataset to significantly advance CMD research.