Continuous systolic and diastolic blood pressure estimation utilizing long short-term memory network

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Cardiovascular Monitoring

Background:

• Chronic diseases require continuous monitoring, including blood pressure (BP).
• Traditional BP estimation methods, often based on Pulse Transit Time (PTT), have limitations.
• Accurate and accessible BP monitoring is crucial for effective disease management.

Purpose of the Study:

• To propose a novel method for blood pressure estimation using Long Short-Term Memory (LSTM) neural networks.
• To introduce and evaluate the Two-stage Zero-order Holding (TZH) algorithm for ambulatory blood pressure (ABP) signal processing.
• To demonstrate the superiority of the proposed method over traditional PTT-based algorithms.

Main Methods:

• Development of a BP estimation model utilizing LSTM recurrent neural networks.
• Implementation of the novel Two-stage Zero-order Holding (TZH) algorithm for ABP signal processing.
• Validation of the method by calculating Root-Mean-Squared Errors (RMSE) between estimated and ground truth BP values.

Main Results:

• The proposed LSTM-based method achieved precise systolic blood pressure (SBP) and diastolic blood pressure (DBP) estimation.
• Average RMSE values were 2.751 mmHg for SBP and 1.604 mmHg for DBP across the sample.
• The TZH algorithm effectively processed ABP signals for enhanced BP estimation.

Conclusions:

• The LSTM-based blood pressure estimation method demonstrates high accuracy and generalization capabilities.
• The proposed methodology offers significant advantages over traditional PTT-based BP estimation techniques.
• This novel approach holds great potential for integration into advanced chronic disease monitoring systems.