Estimating Systolic Blood Pressure Using Convolutional Neural Networks.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Signal Processing

Background:

• Continuous blood pressure (BP) monitoring generates large datasets, crucial for early hypertension and cardiovascular disease (CVD) diagnosis.
• Processing this extensive data for accurate BP estimation presents significant challenges.

Purpose of the Study:

• To develop and compare deep learning techniques for estimating systolic blood pressure (SBP) using electrocardiogram (ECG) and photoplethysmography (PPG) signals.
• To address the data processing challenges in continuous BP monitoring.

Main Methods:

• Two convolutional neural network (CNN) based methods were investigated for SBP estimation.
• Method 1: Employed continuous wavelet transform (CWT) with CNN. Method 2: Utilized raw ECG and PPG signals with CNN and stochastic gradient descent (SGD) optimization.
• The Medical Information Mart for Intensive Care (MIMIC III) database was used, with 70% for training and 30% for testing.

Main Results:

• Both CNN-based methods demonstrated the ability to automatically extract relevant features from ECG and PPG signals, eliminating the need for manual feature engineering.
• High accuracy was achieved by both methods in estimating SBP.
• The proposed CNN architectures are shown to be reliable for SBP estimation.

Conclusions:

• Deep learning, specifically CNNs, provides a powerful and accurate tool for SBP estimation from non-invasive physiological signals.
• These methods offer a reliable and efficient approach to analyzing large datasets from continuous BP monitoring.
• The findings support the potential of these techniques for improved hypertension and CVD management.