Abstract
Cardiovascular diseases remain the leading cause of mortality worldwide; therefore, early-stage noninvasive risk assessment is crucial both for prevention and for the efficient allocation of healthcare resources. Heart rate variability (HRV) is a reliable indicator of autonomic balance, yet in clinical practice it still relies mainly on long-term ECG recordings, which limits large-scale application. Although photoplethysmography (PPG) is widely available, the effectiveness of PPG-derived HRV metrics for detecting ischemic heart disease (IHD) has not been fully established.In this study, PPG signals obtained from a low-cost Zhurek device, combined with machine learning classifiers, enabled IHD detection with an accuracy of 90.82%. Comparison with three-channel Holter ECG showed satisfactory agreement. SHAP and mutual information analyses highlighted the dominant role of frequency-domain features (HF, LF). Data balancing using CTGAN improved training stability.Overall, the results demonstrate that PPG-based HRV analysis offers a feasible, accessible, and interpretable approach for IHD screening. Future steps include multicenter validation, expansion of feature sets, and integration of the method into wearable devices.