[Primary study on recognition of vascular stiffness based on wavelet scattering neural network]

Area of Science:

• Cardiovascular physiology
• Biomedical signal processing
• Machine learning in healthcare

Context:

• Cardiovascular disease is a leading global cause of mortality.
• Arterial stiffness is an independent predictor of cardiovascular diseases, often preceding structural changes.
• Non-invasive methods for assessing vascular stiffness are currently limited.

Purpose:

• To investigate the feasibility of detecting vascular stiffness using Korotkoff signal characteristics.
• To explore the relationship between Korotkoff signal features and vascular compliance.
• To develop a machine learning model for classifying normal versus stiff vessels based on these signals.

Summary:

• Korotkoff signals from normal and stiff vessels were collected and preprocessed.
• Wavelet scattering networks extracted scattering features, which were then used to train a long short-term memory (LSTM) network classifier.
• The classification model achieved 86.4% accuracy, 92.3% sensitivity, and 77.8% specificity in distinguishing between normal and stiff vessels.

Impact:

• The findings suggest that Korotkoff signal characteristics are influenced by vascular compliance.
• This research validates the feasibility of using Korotkoff signals for non-invasive vascular stiffness detection.
• The study proposes a novel approach for non-invasive screening of arterial stiffness, potentially aiding in early cardiovascular disease risk assessment.