Cuffless Blood Pressure Measurement Using Linear and Nonlinear Optimized Feature Selection.

Area of Science:

• Biomedical Engineering
• Physiological Monitoring
• Data Science

Background:

• Cuff-based blood pressure (BP) measurement is invasive and uncomfortable for frequent monitoring.
• Physiological waveform data presents an opportunity for developing noninvasive BP measurement techniques.
• Existing cuffless BP methods often lack robust feature optimization.

Purpose of the Study:

• To propose a novel cuffless, noninvasive blood pressure measurement technique.
• To optimize features from electrocardiogram (ECG) and photoplethysmography (PPG) signals using multivariate symmetric uncertainty (MSU).
• To enhance BP measurement accuracy by considering both linear and nonlinear relationships between physiological signals and BP changes.

Main Methods:

• Utilized multivariate symmetric uncertainty (MSU) for feature selection.
• Employed fast correlation and ReliefF algorithms for feature optimization.
• Applied a penalty-based regression technique to ensure feature relevance and minimize redundancy.
• Validated the technique using the MIMIC-II dataset.

Main Results:

• Achieved a root mean square error (RMSE) of 5.28 mmHg for systolic BP and 5.98 mmHg for diastolic BP.
• Reported mean absolute errors of 4.27 mmHg for SBP and 5.01 mmHg for DBP.
• Met at least grade B criteria according to the British Hypertension Society (BHS) standards for cuffless BP measurement.

Conclusions:

• The proposed cuffless BP measurement technique demonstrates high accuracy and reliability.
• Optimized feature selection using MSU significantly improves noninvasive BP monitoring.
• This method offers a comfortable and frequent alternative to traditional cuff-based BP measurements.