ACNN-BiLSTM: A Deep Learning Approach for Continuous Noninvasive Blood Pressure Measurement Using Multi-Wavelength...

Area of Science:

• Biomedical Engineering
• Physiological Monitoring
• Cardiovascular Health

Background:

• Noninvasive continuous blood pressure (BP) measurement is vital for cardiovascular disease management.
• Single-wavelength Photoplethysmography (PPG) methods face limitations in improving BP prediction accuracy.
• Clinical application and dissemination of current BP monitoring technologies are restricted.

Purpose of the Study:

• To develop and evaluate a novel method for noninvasive continuous BP measurement.
• To enhance BP prediction accuracy by fusing multi-wavelength PPG signals.
• To assess the performance of an attention-based CNN and BiLSTM (ACNN-BiLSTM) model for BP prediction.

Main Methods:

• Proposed a fusion method combining four-wavelength PPG signals with an ACNN-BiLSTM model.
• Processed PPG signals from 162 volunteers to evaluate the fusion method.
• Compared the performance of individual wavelengths versus multi-wavelength fusion using MAE, RMSE, and AAMI criteria.

Main Results:

• The ACNN-BiLSTM model using multi-wavelength PPG fusion achieved superior BP prediction accuracy.
• Achieved a mean absolute error (MAE) ± root mean squared error (RMSE) of 1.67 ± 5.28 mmHg for systolic BP.
• Achieved a MAE ± RMSE of 1.15 ± 2.53 mmHg for diastolic BP.

Conclusions:

• Multi-wavelength PPG fusion combined with the ACNN-BiLSTM model shows significant promise for noninvasive continuous BP monitoring.
• This approach offers improved accuracy compared to traditional single-wavelength PPG methods.
• The developed model could facilitate wider clinical application and dissemination of advanced BP measurement techniques.