End-to-End Deep Learning Architecture for Continuous Blood Pressure Estimation Using Attention Mechanism

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Cardiovascular Monitoring

Background:

• Continuous blood pressure (BP) monitoring is crucial for managing hypertension.
• Existing cuff-less BP monitoring methods often rely on pulse transit time (PTT) and feature extraction.
• There is a need for improved accuracy and efficiency in BP estimation.

Purpose of the Study:

• To propose an end-to-end deep learning architecture for cuff-less BP estimation.
• To leverage raw physiological signals and an attention mechanism for enhanced BP prediction.
• To evaluate the model's performance against state-of-the-art methods.

Main Methods:

• Developed a deep learning model integrating a convolutional neural network, bidirectional gated recurrent unit, and attention mechanism.
• Utilized raw electrocardiogram, ballistocardiogram, and photoplethysmogram signals.
• Employed a subject-specific calibration-based training method.

Main Results:

• The proposed attention-based model demonstrated superior performance in BP estimation compared to models using signal combinations and linear regression with PTT.
• Achieved R² values of 0.52 for systolic BP (SBP) and 0.49 for diastolic BP (DBP).
• Mean absolute errors were 4.06 ± 4.04 mmHg for SBP and 3.33 ± 3.42 mmHg for DBP, complying with global standards.

Conclusions:

• The developed deep learning architecture offers a promising approach for accurate and efficient cuff-less blood pressure monitoring.
• The model's ability to use raw signals and incorporate an attention mechanism enhances its applicability as an analytical metric for BP estimation.
• Results indicate the potential for improved patient management in hypertension through continuous, non-invasive BP tracking.