A Novel CNN-LSTM Model Based Non-Invasive Cuff-Less Blood Pressure Estimation System

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence in Healthcare

Background:

• Continuous blood pressure (BP) monitoring is crucial for cardiovascular health but current methods are invasive or uncomfortable.
• Cuff-based BP monitors apply artificial pressure, causing discomfort, while invasive methods are highly disruptive.
• Existing non-invasive techniques lack the accuracy and convenience needed for widespread clinical and domestic use.

Purpose of the Study:

• To develop a non-invasive, cuff-less, and continuous blood pressure (BP) measurement system.
• To utilize Photoplethysmography (PPG) and Electro-cardiogram (ECG) signals for accurate BP estimation.
• To provide a comfortable and non-disturbing alternative for BP monitoring in various settings.

Main Methods:

• A novel dataset of PPG and ECG signals from 30 healthy participants was created and made publicly available.
• A deep learning model combining a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) layer was designed.
• The CNN-LSTM model was trained and applied to PPG and ECG signals for estimating systolic blood pressure (SBP) and diastolic blood pressure (DBP).

Main Results:

• The proposed CNN-LSTM model achieved a mean absolute error (MAE) of 2.57 mmHg for SBP and 3.44 mmHg for DBP.
• The system demonstrated the lowest error metrics compared to existing studies for cuff-less BP estimation.
• Standard deviation (SD) metrics were comparable to the low MAE, indicating consistent accuracy.

Conclusions:

• The developed CNN-LSTM based system offers a highly accurate and non-invasive solution for continuous blood pressure monitoring.
• This technology has significant clinical relevance for early detection of hypertension and continuous patient monitoring.
• The system enhances patient convenience and accessibility for regular blood pressure self-management.