Principal component analysis-based features generation combined with ellipse models-based classification criterion...

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Ventricular septal defect (VSD) diagnosis relies on accurate heart sound analysis.
• Existing diagnostic methods may lack efficiency or simplicity.
• A novel approach is needed for improved VSD detection.

Purpose of the Study:

• To propose a simple and efficient diagnostic system for VSD.
• To develop a methodology using principal component analysis (PCA) and ellipse models for VSD classification.
• To validate the system's performance against clinical heart sound data.

Main Methods:

• Heart sounds were collected and preprocessed using wavelet decomposition.
• Principal component analysis (PCA) was employed for feature generation from time-frequency matrices.
• Ellipse models, built from support vector machine-based classification curves, defined the VSD diagnosis criterion.

Main Results:

• The proposed system demonstrated high classification accuracy for VSD.
• Accuracies achieved were [Formula: see text] for small VSD, [Formula: see text] for moderate VSD, [Formula: see text] for large VSD, and [Formula: see text] for normal sounds.
• Comparative analysis with other methods confirmed the system's usefulness.

Conclusions:

• The developed diagnostic system offers a simple and efficient method for VSD detection.
• The combination of PCA and ellipse models provides a robust classification criterion.
• This approach shows significant potential for improving the accuracy of VSD diagnosis.