Respiratory sounds classification using statistical biomarker.

Area of Science:

• Medical signal processing
• Biomedical engineering
• Respiratory diagnostics

Background:

• Lung sound (LS) analysis is crucial for diagnosing respiratory conditions.
• Traditional methods may lack the precision needed for accurate classification.
• Feature extraction is key to improving the performance of LS classification models.

Purpose of the Study:

• To propose a new feature extraction technique for lung sound signals.
• To enhance the classification accuracy of respiratory conditions.
• To identify informative features from lung sound signals for diagnostic purposes.

Main Methods:

• Generating intrinsic mode functions (IMFs) from lung sound signals.
• Selecting informative IMFs and extracting features using higher-order moments (mean, standard deviation, skewness, kurtosis).
• Employing an artificial neural network (ANN) classifier to evaluate feature performance.

Main Results:

• The proposed feature extraction method demonstrated superior performance.
• Achieved higher classification accuracy compared to baseline methods.
• Showcased improved sensitivity and specificity in classifying lung sound signals.

Conclusions:

• The novel statistical morphology-based feature extraction technique is effective for lung sound classification.
• The method offers a promising approach for improving the accuracy of respiratory diagnostics.
• Feature vectors derived from higher-order moments of selected IMFs are highly informative.