Continuous blood pressure measurement from one-channel electrocardiogram signal using deep-learning techniques

Area of Science:

• Biomedical Engineering
• Cardiovascular Monitoring
• Signal Processing

Background:

• Continuous blood pressure (BP) monitoring is vital for hypertension management.
• Current methods often require multiple signals (ECG, PPG) or invasive procedures.
• Electrocardiogram (ECG) signals are more accessible via wearable devices.

Purpose of the Study:

• To develop and validate a novel continuous BP estimation method using single-channel ECG signals.
• To enable unobtrusive and accessible BP monitoring.
• To assess the model's performance against established clinical standards.

Main Methods:

• A deep learning model fusing residual networks and long short-term memory was developed.
• The model processes spatial-temporal information from ECG signals for BP estimation.
• The model was trained and validated using the public multiparameter intelligent monitoring waveform database.

Main Results:

• The model achieved estimation errors compliant with Association for the Advancement of Medical Instrumentation standards for mean arterial pressure (MAP) and diastolic BP (DBP).
• BP estimation achieved Grade A for MAP and DBP, and Grade B for systolic BP (SBP) according to British Hypertension Society standards.
• Validation on an independent dataset of arrhythmia patients demonstrated high accuracy and feasibility.

Conclusions:

• Continuous BP estimation is feasible using only one-channel ECG signals.
• The proposed method offers a promising approach for ubiquitous, non-invasive BP monitoring.
• This technology has significant potential for widespread healthcare applications and early hypertension detection.