A system for heart sounds classification.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Non-invasive cardiac disease diagnosis is crucial for public health.
• Existing automated heart sound analysis methods struggle with signal complexity and variety.
• Advancements in phonocardiography signal quality necessitate improved diagnostic algorithms.

Purpose of the Study:

• To develop an advanced heart sound classification technique for automated cardiac diagnostic systems.
• To address the challenges posed by non-stationary phonocardiography signals and diverse pathological heart sounds.
• To enhance the accuracy, reduce complexity, and broaden the scope of heart disease identification.

Main Methods:

• Feature extraction using Linear Predictive Coding (LPC) coefficients.
• Development of a hybrid classifier combining Support Vector Machine (SVM) and Modified Cuckoo Search (MCS) algorithm.
• Evaluation of the system's performance against established classification methods.

Main Results:

• The developed system achieved over 93% accuracy in classifying twelve different heart sound classes simultaneously.
• Demonstrated superior performance compared to four other major classification techniques.
• Showcased improvements in accuracy, complexity, and the range of distinguishable heart sounds.

Conclusions:

• The proposed heart sound classification technique offers a reliable and efficient solution for automated cardiac diagnostics.
• The combination of LPC and SVM-MCS provides a robust approach to analyzing complex phonocardiography signals.
• This method holds significant potential for improving early detection and management of cardiac diseases.