Combined deep CNN-LSTM network-based multitasking learning architecture for noninvasive continuous blood pressure...

Area of Science:

• Biomedical Engineering
• Machine Learning in Healthcare
• Cardiovascular Monitoring

Background:

• Pulse Arrival Time (PAT), derived from ECG and PPG signals, is a key indicator for noninvasive blood pressure estimation.
• Inconsistent signal shapes due to physiological variations complicate accurate PAT measurement and blood pressure estimation.
• Existing methods often require complex preprocessing, limiting practical application.

Purpose of the Study:

• To develop a noninvasive, continuous blood pressure estimation algorithm using a novel feature derived from ECG and PPG signals.
• To propose a deep CNN-LSTM-based multitasking model for simultaneous systolic (SBP) and diastolic (DBP) blood pressure prediction.
• To validate the algorithm's accuracy against established clinical guidelines.

Main Methods:

• A deep convolutional neural network (CNN) combined with a long short-term memory (LSTM) network was employed as a multitasking machine learning model.
• The algorithm utilizes the difference between ECG and PPG signals as a new feature incorporating PAT information.
• Data from 48 patients (38 for training, 10 for testing) on the PhysioNet database were used for model development and validation.

Main Results:

• The model achieved high prediction accuracies for SBP (0.0 ± 1.6 mmHg) and DBP (0.2 ± 1.3 mmHg).
• These results demonstrate the algorithm's potential for accurate blood pressure monitoring.
• The performance satisfied the stringent requirements of the BHS, AAMI, and IEEE standards for blood pressure measurement devices.

Conclusions:

• The proposed deep learning algorithm offers a promising noninvasive and continuous method for blood pressure estimation.
• Utilizing the ECG-PPG difference as a feature simplifies preprocessing and enhances accuracy.
• The model's adherence to international standards suggests its clinical viability for blood pressure monitoring.