A multimodal dataset to explore the potential biomarker for coronary microvascular disease

Coronary microvascular disease (CMD), particularly prevalent among women, is associated with increased morbidity and mortality. Clinical screening for CMD is crucial for effective management stratification. However, the scarcity of publicly available screening-level data poses a significant challenge to biomarker discovery. In this study, we prospectively enrolled 80 female patients with angina but no obstructive coronary artery disease and 39 age-matched female controls. All participants underwent adenosine stress testing with continuous electrocardiogram (ECG) monitoring across Rest, Stress, and Recovery stages. CMD diagnosis was confirmed by coronary flow reserve (CFR) <2.0, as determined by PET/CT imaging during the Stress stage. Using ECG variables from different stages, we developed five machine learning models to predict CMD, finding that ECG data from various stages provide valuable information for CMD classification.To further validate the diagnostic utility of ECG in distinguishing CMD from diseases with similar clinical presentations, such as mental stress-induced myocardial ischemia (MSIMI), the same cohort underwent mental stress stimulation, with ECG recorded across three phases, and MSIMI diagnosis confirmed by a summed differential score (SDS) ≥3 obtained from PET/CT. We then assessed the correlation between ECG variables and the diagnostic criteria (CFR for CMD and SDS for MSIMI), identifying two disease-specific sets of ECG variables. Our data demonstrate the utility of multi-stage ECG in CMD screening and preliminary differentiation. We anticipate that this new dataset will significantly advance CMD research.