Robust blood pressure estimation from finger photoplethysmography using age-dependent linear models

Area of Science:

• Biomedical Engineering
• Physiological Monitoring
• Signal Processing

Background:

• Accurate blood pressure (BP) monitoring is crucial for cardiovascular health management.
• Existing methods often require invasive procedures or specialized equipment.
• Photoplethysmography (PPG) offers a non-invasive approach for physiological monitoring.

Purpose of the Study:

• To develop an efficient and robust age-dependent multiple linear regression (MLR) model for BP estimation.
• To utilize single-source PPG signals and biometrics for BP estimation.
• To enable integration into microcontroller-based pulse oximeters.

Main Methods:

• Extracted hemodynamic features from PPG waveforms, derivatives, and biometrics.
• Evaluated whole-based, feature-based, and fusion models across different age groups.
• Employed leave-one-subject-out cross-validation on 1086 subjects.
• Investigated the impact of biometrics and long-term calibration effects.

Main Results:

• Fusion models excelled with well-defined PPG signals; feature-based models performed better with damped signals.
• Age significantly improved BP estimation accuracy for older subjects (>50 years), particularly with well-defined PPG features.
• Age had minimal impact on BP estimation for younger subjects (≤50 years).
• Subject-specific calibration reduced errors and improved linearity.

Conclusions:

• Age-dependent MLR models provide reasonable BP estimation accuracy.
• Tailoring models based on age and signal quality is essential, especially for diastolic BP in older adults.
• This approach can enhance the functionality of existing pulse oximeters.