Real-Time Cuffless Continuous Blood Pressure Estimation Using Deep Learning Model.

Area of Science:

• Biomedical Engineering
• Health Informatics
• Artificial Intelligence in Medicine

Background:

• Continuous blood pressure monitoring is crucial for early detection of cardiovascular diseases.
• Real-time estimation of blood pressure can significantly improve patient outcomes and reduce mortality.
• Existing methods for blood pressure monitoring have limitations in continuous, non-invasive assessment.

Purpose of the Study:

• To develop and evaluate deep learning regression models for real-time estimation of systolic blood pressure (SBP) and diastolic blood pressure (DBP).
• To utilize electrocardiogram (ECG) and photoplethysmogram (PPG) signals for non-invasive blood pressure monitoring.
• To compare the proposed deep learning models against traditional machine learning and existing deep learning approaches.

Main Methods:

• Proposed a deep learning architecture incorporating bidirectional Long Short-Term Memory (LSTM) layers with residual connections.
• Utilized ECG and PPG signals from the Physionet MIMIC II dataset.
• Compared model performance against traditional machine learning and a benchmark deep learning model.

Main Results:

• The proposed deep learning model demonstrated superior performance in estimating SBP and DBP compared to existing methods.
• Achieved accurate real-time blood pressure estimation, indicating significant potential for clinical utility.
• The model's effectiveness was validated using a comprehensive intensive care unit dataset.

Conclusions:

• Deep learning models, particularly the proposed LSTM architecture, offer a promising approach for accurate, real-time, non-invasive blood pressure estimation.
• The findings suggest the potential for effective clinical application of these models in patient monitoring.
• Further research and validation are warranted for widespread clinical adoption.