A texture-based classification of crackles and squawks using lacunarity

Area of Science:

• Bioacoustics
• Medical Signal Processing
• Pulmonary Medicine

Background:

• Discontinuous breath sounds (DBSs) include fine crackles (FCs), coarse crackles (CC), and squawks (SQ).
• Accurate classification of DBSs is crucial for diagnosing pulmonary pathologies.
• Existing methods may lack efficiency or fail to capture subtle acoustic differences.

Purpose of the Study:

• To present an automatic classification method, LAC, for discriminating between FCs, CCs, and SQs.
• To introduce a texture-based approach using lacunarity for DBS analysis.
• To evaluate the efficiency and accuracy of LAC in classifying DBS categories.

Main Methods:

• Wavelet-based denoising to remove background noise from DBS.
• Lacunarity analysis at an optimum scale to capture texture.
• Modeling lacunarity trajectory across scales using a three-parameter hyperbola.
• Application of the LAC method to 363 DBS from four lung sound databases.

Main Results:

• LAC achieved high classification accuracies: FC-CC (100%), FC-SQ (100%), CC-SQ (99.62%-100%), and FC-CC-SQ (99.75%-100%).
• The method demonstrates high accuracy without significant computational complexity.
• LAC introduces a 'texture' concept, aligning with physician perception of bioacoustic signals.

Conclusions:

• LAC offers an efficient and accurate automatic classification of DBSs.
• The texture-based approach provides a novel perspective on pulmonary acoustical changes.
• LAC's simplicity allows for potential real-time implementation in clinical decision support systems.