From Pulse to Insight: An End-To-End Explainable Deep Spatio-Temporal Fusion Network for Cuffless Blood Pressure...

Area of Science:

• Biomedical Engineering
• Cardiovascular Technology
• Artificial Intelligence in Healthcare

Background:

• Continuous blood pressure monitoring is crucial for cardiovascular disease management.
• Traditional cuff-based methods are invasive and inconvenient for long-term use.
• Photoplethysmography (PPG) signals offer a potential non-invasive alternative.

Purpose of the Study:

• To develop and validate a novel cuffless blood pressure estimation system using only PPG signals.
• To integrate deep learning for robust feature extraction and accurate blood pressure prediction.
• To enhance clinical decision-making through an interpretable AI model.

Main Methods:

• Utilized a deep spatio-temporal fusion network with convolutional and recurrent layers.
• Trained and validated the model on the PulseDB dataset.
• Employed Grad-Cam for model explainability and developed a custom PPG acquisition device.

Main Results:

• Achieved mean absolute errors of 3.57 mmHg (systolic) and 2.20 mmHg (diastolic).
• Demonstrated low bias and acceptable variability, meeting AAMI and BHS validation standards.
• The system provides reliable, non-invasive, and interpretable cuffless blood pressure monitoring.

Conclusions:

• The proposed system offers a reliable, non-invasive, and interpretable solution for continuous blood pressure monitoring.
• This technology has significant potential to improve early detection and management of hypertension and cardiovascular risks.
• Explainability of the AI model enhances trust and clinical applicability.