GHC-net: A gramian angular field based hybrid CNN for cuffless blood pressure classification using PPG signals

Area of Science:

• Biomedical Engineering
• Cardiovascular Disease Research
• Signal Processing

Background:

• Cuffless blood pressure (BP) monitoring using photoplethysmography (PPG) is crucial for cardiovascular disease prevention.
• Traditional temporal feature extraction methods have limitations in capturing complex PPG signal dynamics.

Purpose of the Study:

• To propose a novel encoding strategy for cuffless BP classification.
• To enhance the accuracy and efficiency of BP monitoring through advanced deep learning techniques.

Main Methods:

• Transformed 1D PPG signals into 2D matrices using Gramian Angular Field (GAF) coding.
• Developed a deep neural network, GHC-Net, incorporating hybrid dilation convolution (HDC) and Remaining Effective Channel Attention (RECA) modules.
• Utilized GAF-encoded PPG data for triple BP classification (normotensive, prehypertensive, hypertensive).

Main Results:

• Achieved 76.39% accuracy and 76.68% F1 score on the public PPG-BP dataset.
• Attained 92.66% accuracy and 93.33% F1 score on a private dataset.
• Demonstrated the model's ability to capture multi-scale spatial-temporal correlations and cross-channel dependencies.

Conclusions:

• The proposed GAF coding and GHC-Net offer an efficient and accurate automated BP classification method.
• This technique shows significant potential for home care applications in early hypertension detection and monitoring.