PPG2ABP: Translating Photoplethysmogram (PPG) Signals to Arterial Blood Pressure (ABP) Waveforms

Area of Science:

• Biomedical Engineering
• Cardiovascular Health
• Artificial Intelligence in Medicine

Background:

• Cardiovascular diseases are a leading cause of global mortality.
• Continuous blood pressure monitoring is crucial but often invasive and unreliable with current non-invasive methods.
• Existing non-invasive techniques for blood pressure estimation face limitations with signal quality and feature extraction.

Purpose of the Study:

• To develop a non-invasive method for estimating continuous arterial blood pressure (ABP) waveforms using Photoplethysmogram (PPG) signals.
• To leverage deep learning to overcome limitations of handcrafted features and signal variability in existing approaches.
• To improve the accuracy and reliability of non-invasive blood pressure monitoring.

Main Methods:

• A two-stage cascaded deep learning model, PPG2ABP, was developed.
• The model estimates continuous ABP waveforms directly from PPG signals.
• No explicit training for specific blood pressure metrics (DBP, MAP, SBP) was performed.

Main Results:

• The PPG2ABP model achieved a mean absolute error of 4.604 mmHg for ABP waveform estimation, preserving shape, magnitude, and phase.
• Estimated Diastolic Blood Pressure (DBP), Mean Arterial Pressure (MAP), and Systolic Blood Pressure (SBP) outperformed existing methods.
• Achieved Grade A in the British Hypertension Society (BHS) Standard for DBP and MAP, and satisfied the Association for the Advancement of Medical Instrumentation (AAMI) standard.

Conclusions:

• The PPG2ABP deep learning method offers a promising non-invasive approach for continuous blood pressure monitoring.
• The model demonstrates superior accuracy in estimating ABP waveforms and key blood pressure parameters.
• The results indicate potential for improved cardiovascular disease management through advanced non-invasive monitoring.