Using Bayesian Optimization and Wavelet Decomposition in GPU for Arterial Blood Pressure Estimation.

Area of Science:

• Biomedical Engineering
• Machine Learning in Healthcare
• Signal Processing

Background:

• Continuous arterial blood pressure (ABP) monitoring is crucial for patient care but often invasive, posing risks.
• Existing noninvasive methods may lack efficiency in model optimization.

Purpose of the Study:

• To develop and evaluate a noninvasive methodology for optimizing ABP estimators.
• To leverage machine learning and GPU acceleration for efficient ABP estimation.

Main Methods:

• Utilized electrocardiogram (ECG) and photoplethysmography (PPG) signals.
• Employed an XGBoost machine learning model optimized with Bayesian techniques.
• Leveraged Graphics Processing Unit (GPU) for accelerated computation.
• Validated the methodology using the MIMIC-III Waveform Database.

Main Results:

• Achieved mean absolute errors of 15.85 mmHg for systolic and 11.59 mmHg for diastolic pressures.
• Demonstrated performance comparable to state-of-the-art methods.
• Significantly reduced computational time for model optimization.

Conclusions:

• The proposed noninvasive methodology offers an efficient and automated approach to ABP estimation.
• This technique enhances patient safety by replacing invasive monitoring methods.
• The computational efficiency makes it suitable for real-time clinical applications.