SleepBP-Net: A Time-Distributed Convolutional Network for Nocturnal Blood Pressure Estimation from...

Area of Science:

• Biomedical Engineering
• Cardiovascular Health Monitoring
• Artificial Intelligence in Healthcare

Background:

• Nocturnal blood pressure (BP) monitoring is crucial for cardiovascular health assessment.
• Current methods for overnight BP measurement are often invasive or burdensome.
• There is a need for non-invasive, cuffless technologies for continuous sleep BP monitoring.

Purpose of the Study:

• To investigate the estimation of nocturnal BP from overnight photoplethysmography (PPG) signals.
• To develop and validate a novel machine learning model for long-term nocturnal BP estimation.
• To assess the efficacy of a lightweight convolutional recurrent network for unobtrusive BP monitoring during sleep.

Main Methods:

• Development of SleepBP-Net, a time-distributed convolutional recurrent network, for nocturnal BP estimation.
• Utilized overnight PPG signals collected in a clinical sleep laboratory setting.
• Validated the model using Portapres BP measurements as a reference, employing leave-one-subject-out cross-validation.

Main Results:

• Initial model achieved a Mean Absolute Error (MAE) of 15.7 mmHg (SBP) and 12.1 mmHg (DBP) for 1-minute PPG sequences.
• Personalized model significantly improved accuracy to 7.8 mmHg (SBP) and 5.9 mmHg (DBP).
• Extending sequence length to 30 minutes further reduced MAE to 7.2 mmHg (SBP) and 5.7 mmHg (DBP), demonstrating the value of long-term patterns.

Conclusions:

• SleepBP-Net effectively estimates nocturnal BP from PPG signals by leveraging long-term temporal patterns.
• Hybrid convolutional recurrent networks show superior performance over purely convolutional approaches for this task.
• The proposed model offers a promising solution for unobtrusive, efficient, and accurate real-world nocturnal BP monitoring.