Hybrid CNN-LSTM Model for Evaluating Heart Rate Variability from Pulse-to-Pulse Intervals

Heart rate variability (HRV) is a key marker of autonomic function and cardiovascular health. This study proposes a hybrid CNN-LSTM model to estimate ECG-derived R-R intervals (RRI) from Photoplethysmography (PPG) signals, enabling non-invasive HRV monitoring. Data from 19 subjects under controlled breathing conditions (6, 18, and 30 breaths per minute) were used to train and validate the model. Results show a high correlation between predicted and actual RRI, with mean absolute error (MAE) decreasing from 62.01 ms to 49.62 ms across subjects. While the model accurately captures HRV at lower breathing rates, frequency domain analysis reveals systematic underestimation of very low frequency (VLF), low frequency (LF), and high frequency (HF) components (p < 0.01). These findings highlight the potential of deep learning for real-time, non-invasive HRV assessment.Clinical Relevance- This study validates a deep learning approach for estimating HRV from PPG signals, offering a noninvasive alternative to ECG. The model provides accurate RRI predictions, particularly at lower breathing rates, enabling continuous cardiovascular monitoring via wearable devices for remote patient care and early disease detection.