Automated Diagnosis of Heart Sounds Using Rule-Based Classification Tree

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• Accurate heart sound analysis is crucial for diagnosing cardiac conditions.
• Automated methods can improve the efficiency and consistency of heart sound diagnosis.

Purpose of the Study:

• To develop an automated method for classifying heart status using heart sound signals.
• To differentiate between normal heart sounds and specific abnormalities: aortic valve stenosis, aortic insufficiency, and ventricular septum defect.

Main Methods:

• Automatic detection and segmentation of heart sound signal cycles.
• Preprocessing using discrete wavelet transform and calculation of largest Lyapunov exponents for feature extraction.
• Classification using a rule-based decision tree based on extracted dynamical features.

Main Results:

• The developed automated method successfully classified heart sound signals.
• Achieved a high success rate of 95.5% on twenty-two datasets of normal and abnormal heart sounds.
• Demonstrated that classification rules can be modified to further improve diagnostic accuracy.

Conclusions:

• The proposed rule-based classification tree offers an effective automated approach for heart sound diagnosis.
• Largest Lyapunov exponents derived from heart sound time series are valuable dynamical features for classification.
• The method shows potential for improving the accuracy and accessibility of automated cardiac diagnosis.