Continuous Patient-Independent Estimation of Respiratory Rate and Blood Pressure Using Robust Spectro-Temporal...

Area of Science:

• Biomedical Engineering
• Physiological Signal Processing
• Medical Informatics

Background:

• Continuous monitoring of vital signs is crucial for patient care.
• Photoplethysmogram (PPG) signals offer a non-invasive method for vital sign estimation.
• Existing PPG-based methods often suffer from patient-specific variations and artifacts.

Purpose of the Study:

• To develop a patient-independent approach for continuous estimation of respiration rate (RR) and blood pressure (BP) from PPG signals.
• To extract robust spectro-temporal features from PPG signals that are independent of individual patient morphology.
• To improve the accuracy and reliability of non-invasive vital sign monitoring.

Main Methods:

• PPG signal pre-processing using Incremental Merge Segmentation with adaptive thresholding to remove artifacts.
• Extraction of multiple parameters, novel spectral and statistical features from cleaned PPG for RR and BP estimation.
• Correlation-based feature selection and neural network regression for accurate RR and BP measurements.

Main Results:

• The proposed method achieved state-of-the-art performance in estimating both RR and BP.
• RR estimation resulted in low mean absolute errors on CapnoBase and BIDMC datasets (0.4 and 0.5, respectively).
• BP estimation met AAMI and BHS Class A criteria with low errors (Systolic: 5.0±6.3, Diastolic: 3.0±4.0) on a large MIMIC-II dataset.

Conclusions:

• The developed patient-independent approach provides reliable vital sign estimates from PPG signals.
• Robust feature extraction and selection methods effectively reduce patient dependency.
• This technique offers a promising solution for continuous, non-invasive health monitoring.