A novel intelligent system based on adjustable classifier models for diagnosing heart sounds.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiovascular Diagnostics

Background:

• Accurate diagnosis of heart sounds (HSs) is crucial for cardiovascular health assessment.
• Existing methods for HS analysis often lack precision in segmentation and feature extraction.
• The development of intelligent systems can improve the objectivity and accuracy of HS diagnosis.

Purpose of the Study:

• To propose a novel intelligent diagnostic system for automated heart sound analysis.
• To introduce innovative techniques for automatic segmentation, feature extraction, and classification of HSs.
• To enhance diagnostic accuracy for various heart conditions through adjustable classifier models.

Main Methods:

• Automatic segmentation and extraction of first and second heart sounds using the short-time modified Hilbert transform (STMHT).
• Extraction of secondary envelope-based diagnostic features and frequency features using novel methods and thresholding.
• Application of Principal Component Analysis (PCA) for feature reduction and Gaussian Mixture Models (GMM) with adjustable confidence levels for classification.

Main Results:

• The system successfully segmented and extracted key components of heart sounds.
• Novel envelope-based features and frequency features were automatically extracted.
• The system achieved high classification accuracies for various heart conditions, including MR (99.67%) and VSD (99.91%), by optimizing confidence levels.

Conclusions:

• The proposed intelligent system demonstrates significant potential for accurate and automated heart sound diagnosis.
• The integration of advanced signal processing and adaptive machine learning enhances diagnostic performance.
• This system offers a promising tool for clinical application in cardiovascular disease detection.