An Irregularity Measurement Based Cardiac Status Recognition Using Support Vector Machine.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Phonocardiogram (PCG) analysis is crucial for diagnosing heart conditions.
• Environmental noise and lung sounds often interfere with PCG signal interpretation.
• Existing methods struggle with signal quality degradation.

Purpose of the Study:

• To develop an automated, robust feature extraction technique for PCG analysis.
• To assess the effectiveness of sample entropy for quantifying heart sound complexity.
• To evaluate the proposed method's performance under noisy conditions and compare it with conventional approaches.

Main Methods:

• Utilized a nonlinear signal processing framework to estimate heart sound complexity via sample entropy.
• Employed a support vector machine (SVM) classifier to evaluate feature effectiveness.
• Tested the framework on a dataset of 120 individuals (60 healthy, 60 pathological) across various Signal-to-Noise Ratio (SNR) levels (-5 to 10 dB).

Main Results:

• The sample entropy-based method achieved high classification accuracy, reaching 96.67% for clean data and 91.66% for noisy data (SNR 10 dB).
• Performance accuracy in noisy conditions remained comparable to that of clean signals.
• The proposed method demonstrated superior performance compared to conventional waveform, spectral, and spectrogram analyses.

Conclusions:

• The proposed automated feature extraction technique based on sample entropy is robust and effective for PCG analysis.
• This method significantly improves heart sound analysis accuracy, particularly in the presence of noise and interference.
• The findings suggest a promising advancement for automated cardiac auscultation and diagnosis.