Features Extraction for Cuffless Blood Pressure Estimation by Autoencoder from Photoplethysmography.

Area of Science:

• Biomedical Engineering
• Cardiovascular Physiology
• Machine Learning

Background:

• Cuffless blood pressure (BP) estimation using Photoplethysmograph (PPG) sensors is an active research area.
• Current machine learning methods rely on limited PPG waveform features, leading to inadequate accuracy for medical applications.

Purpose of the Study:

• To develop an advanced method for cuffless BP estimation by leveraging autoencoders to extract comprehensive features from PPG signals.
• To improve the accuracy of BP estimation beyond existing techniques.

Main Methods:

• Utilized an autoencoder to extract 100 complex features from the PPG signal and its first and second-order derivatives.
• Applied the autoencoder model to estimate BP from PPG data.

Main Results:

• The autoencoder-based method improved the correlation coefficient (R) from 0.67 to 0.78.
• Reduced the standard deviation (SD) of the difference between measured and estimated BP from 13.97 to 11.86.
• These improvements were validated across 1363 test subjects.

Conclusions:

• Autoencoders can effectively extract intricate features from PPG waveforms for more accurate cuffless BP estimation.
• The proposed method shows significant potential for advancing non-invasive BP monitoring technology.