A two-branch framework for blood pressure estimation using photoplethysmography signals with deep learning and...

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Cardiovascular Monitoring

Background:

• Accurate blood pressure (BP) monitoring is crucial for cardiovascular health management.
• Current cuff-based methods are invasive and limit continuous monitoring.
• Photoplethysmography (PPG) offers a non-invasive alternative for BP estimation.

Purpose of the Study:

• To develop and validate a novel dual-branch deep learning framework for precise, cuffless blood pressure estimation using PPG signals.
• To incorporate clinical prior knowledge and model diurnal variations in blood pressure.
• To achieve accurate estimation of systolic blood pressure (SBP) and diastolic blood pressure (DBP).

Main Methods:

• A dual-branch framework was designed, utilizing deep PPG features from MobileViTv2 and Vgg19 backbones and multi-dimensional waveform parameters.
• Features from both branches were fused, considering diurnal BP variations and employing AutoML for period-specific SBP/DBP models.
• The model was trained on the HRSD dataset and validated on the MIMIC-IV dataset.

Main Results:

• Achieved Mean Absolute Errors (MAE) as low as 2.65 mmHg (SBP) and 2.56 mmHg (DBP) in the evening on the HRSD dataset.
• Demonstrated strong generalization performance on MIMIC-IV with MAEs of 4.34 mmHg (SBP) and 3.11 mmHg (DBP).
• Met Association for the Advancement of Medical Instrumentation standards and achieved Grade A of British Hypertension Society standards.

Conclusions:

• The proposed framework provides an accurate and reliable non-invasive blood pressure monitoring technology.
• This method is suitable for continuous health monitoring and cardiovascular disease prevention.
• The integration of deep learning and clinical knowledge enhances cuffless BP estimation accuracy.