CHAMPS: Cardiac health Hypergraph Analysis using Multimodal Physiological Signals.

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Non-invasive screening for Coronary Artery Disease (CAD) is crucial for early detection and management.
• Existing methods utilize various physiological signal features but often struggle to capture complex interrelationships.

Purpose of the Study:

• To propose a novel hypergraph-based approach for representing and analyzing features from multiple physiological signals for CAD screening.
• To enhance the classification accuracy of CAD using a hypergraph formalism.

Main Methods:

• Features from Photoplethysmogram (PPG) and Phonocardiogram (PCG) signals were extracted and interconnected using hyperedges.
• Metadata features (age, weight, height) were integrated into the hypergraph structure.
• Hypergraph Laplacian was employed as a derived feature for classification between CAD and non-CAD.

Main Results:

• The proposed hypergraph method demonstrated high performance in classifying Coronary Artery Disease.
• Achieved 98% Sensitivity, 82% Specificity, and 92% overall classification accuracy.
• Validation was performed on PPG and PCG data collected in a clinical setting.

Conclusions:

• Hypergraph formalism offers a flexible and powerful way to capture intricate feature interrelationships for medical diagnosis.
• The developed method shows significant promise for accurate and non-invasive Coronary Artery Disease screening.