CHANGE: Cardiac Health Analysis Using Graph Eigenvalues.

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Coronary Artery Disease (CAD) is a major cause of mortality.
• Photoplethysmogram (PPG) signals offer non-invasive features for CAD classification.
• Existing methods have limitations in exploiting feature dependencies.

Purpose of the Study:

• To develop a novel graph-based method for Coronary Artery Disease (CAD) classification.
• To leverage dependencies between Photoplethysmogram (PPG) and metadata features.
• To improve the accuracy of CAD detection using non-invasive signals.

Main Methods:

• Representing cardiac health as a Cardiac Health Graph (CHG).
• Computing spectral features from the eigenvalues of the CHG Laplacian.
• Employing k-means clustering for CAD and non-CAD classification.

Main Results:

• Achieved 88% accuracy in unsupervised classification of CAD.
• Demonstrated superior performance compared to baseline and state-of-the-art methods.
• Successfully exploited feature dependencies using the graph formulation.

Conclusions:

• The proposed Cardiac Health Graph (CHG) formulation is effective for CAD detection.
• Graph-based spectral features enhance the classification of cardiac health from PPG signals.
• This method shows promise for non-invasive CAD diagnosis.