An Adaptive Weight Learning-Based Multitask Deep Network for Continuous Blood Pressure Estimation Using Electrocardiogram Signals | JoVE Visualize

Area of Science:

Biomedical Engineering
Artificial Intelligence in Healthcare
Cardiovascular Monitoring

Background:

Continuous blood pressure monitoring is crucial for managing hypertension.
Current wearable devices often lack multi-signal acquisition capabilities due to constraints.
Electrocardiogram (ECG) and photoplethysmography (PPG) signals offer potential for non-invasive blood pressure estimation.

Purpose of the Study:

To propose a novel deep learning framework for continuous blood pressure estimation using only single-lead ECG signals.
To develop an adaptive weight learning scheme for optimizing multitask learning in blood pressure estimation.
To validate the performance of the proposed model against established medical standards.

Main Methods:

Utilized a 2-layer bidirectional long short-term memory (LSTM) network as a shared feature extractor.
Implemented three identical 2-layer fully connected networks for task-specific blood pressure estimation (systolic, diastolic, mean arterial).
Employed an adaptive weight learning strategy based on validation loss trends to automatically adjust task importance.

Main Results:

Achieved excellent performance in estimating systolic blood pressure (0.12±10.83 mmHg), diastolic blood pressure (0.13±5.90 mmHg), and mean arterial pressure (0.08±6.47 mmHg) on the MIMIC-II database.
The model's accuracy significantly exceeded the requirements set by the British Hypertension Society and US Association for the Advancement of Medical Instrumentation.
Demonstrated the feasibility of accurate blood pressure estimation using solely ECG signals.

Conclusions:

The proposed adaptive weight learning-based multitask deep learning framework enables accurate continuous blood pressure estimation from single-lead ECG signals.
This approach overcomes the limitations of current wearable devices by relying on a single, easily acquired physiological signal.
Deployment in healthcare systems with wearable ECG devices can facilitate long-term blood pressure monitoring, potentially reducing hypertension-related complications.