An Estimation Method of Continuous Non-Invasive Arterial Blood Pressure Waveform Using Photoplethysmography: A U-Net...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiovascular Physiology

Background:

• Non-invasive blood pressure (BP) monitoring is crucial for managing hypertension and cardiovascular diseases.
• Photoplethysmogram (PPG) signals offer a non-invasive method for BP assessment, driving recent research.
• Accurate and accessible BP monitoring remains a significant clinical challenge.

Purpose of the Study:

• To develop and validate a deep learning model for non-invasive estimation of arterial BP (ABP) waveforms using PPG signals.
• To accurately derive systolic BP (SBP), diastolic BP (DBP), and mean arterial pressure (MAP) from estimated ABP waveforms.
• To assess the performance of the proposed method against established clinical standards.

Main Methods:

• A U-net deep learning architecture was employed, utilizing fingertip PPG signals as input.
• The model was trained and evaluated on data from 100 subjects across the MIMIC and MIMIC-III databases.
• Performance was quantified using Pearson's correlation coefficient and mean absolute error (MAE) for SBP, DBP, and MAP.

Main Results:

• The predicted ABP waveforms demonstrated a high correlation (0.993) with reference waveforms.
• Mean absolute errors were 3.68 ± 4.42 mmHg for SBP, 1.97 ± 2.92 mmHg for DBP, and 2.17 ± 3.06 mmHg for MAP.
• The results met the AAMI standard and achieved Grade A under the BHS standard.

Conclusions:

• The proposed U-net deep learning method provides an efficient and accurate approach for non-invasive ABP waveform estimation from fingertip PPG.
• This technique holds potential for improved BP monitoring in clinical and remote settings.
• The high accuracy and adherence to clinical standards suggest clinical viability for the developed model.