A Deep Convolution Method for Hypertension Detection from Ballistocardiogram Signals with Heat-Map-Guided Data...

Area of Science:

• Biomedical Engineering
• Cardiovascular Health
• Artificial Intelligence in Medicine

Background:

• Hypertension (HPT) is a major risk factor for stroke, coronary artery disease, and heart failure.
• Non-contact detection and continuous monitoring of HPT are challenging clinical needs.
• Ballistocardiography (BCG) signals, reflecting heartbeat-induced body motion, offer potential for HPT assessment.

Purpose of the Study:

• To develop an end-to-end deep convolutional model (BH-Net) for hypertension detection using BCG signals.
• To propose a data augmentation scheme to improve the accuracy of HPT detection from BCG.
• To evaluate the proposed model and data augmentation against existing state-of-the-art methods.

Main Methods:

• An end-to-end deep convolutional neural network, termed BH-Net, was designed for HPT detection.
• A novel data augmentation technique focused on J-peak neighborhoods within BCG time sequences was implemented.
• The BH-Net model and data augmentation scheme were rigorously evaluated using a public BCG dataset.

Main Results:

• The proposed BH-Net model achieved an average accuracy of 97.93% and an average F1-score of 97.62%.
• The model significantly outperformed existing state-of-the-art methods for HPT detection using BCG.
• The data augmentation scheme demonstrably improved the performance of both traditional machine learning and comparative deep learning models.

Conclusions:

• BH-Net represents a highly effective deep learning approach for non-contact hypertension detection via BCG signals.
• The proposed data augmentation strategy enhances the robustness and accuracy of BCG-based HPT detection.
• This research offers a promising, non-invasive tool for hypertension screening and monitoring.