Estimation of Blood Pressure Response to Physiological Maneuvers in Hypertensive Patients using Bioimpedance

Blood pressure (BP) fluctuations arise from various physiological states, yet current cuffless BP estimation methods fail to account for these dynamic factors, limiting accuracy and interpretability. Existing models rely on wearable sensor data, such as bioimpedance (Bio-Z) and photoplethysmography (PPG), but struggle with indirect BP relationships, significant inter- and intra-patient variability, and the lack of physiological context in model training. We propose a physiology-aware BP estimation framework that explicitly incorporates physiological state information (e.g., stress, exercise) to improve prediction accuracy and robustness. Our approach extracts physiology-specific features, dynamically integrates multiple estimators, and enhances generalization to unseen physiological conditions. Our approach has achieved an improvement of 16.60% in RMSE for systolic BP (SBP) and 14.04% for diastolic BP (DBP), resulting in an RMSE of 7.58 mmHg for SBP and 4.94 mmHg for DBP on a dataset comprising 39 hypertensive participants. By addressing key limitations in current methods, our framework advances the reliability of cuffless BP monitoring while providing critical insights into the physiological mechanisms driving BP changes.