Construction and Validation of an Automatic Segmentation Method for Respiratory Sound Time Labels.

Area of Science:

• Pulmonary Medicine
• Biomedical Engineering
• Medical Acoustics

Background:

• Auscultation is vital for diagnosing respiratory diseases but is subjective.
• Personal experience and environmental factors can lead to diagnostic errors in auscultation.
• Accurate analysis of respiratory sounds aids in diagnosing and treating respiratory conditions.

Purpose of the Study:

• To develop an analytical method for visualizing and digitizing respiratory audio data.
• To validate the method's ability to differentiate between various background respiratory diseases.

Main Methods:

• Collected respiratory sounds from 84 patients using an electronic stethoscope in a quiet environment.
• Segmented audio data, distinguishing heart and respiratory sounds, and identifying inspiratory/expiratory phases.
• Utilized a custom tool for automatic segmentation encoding and feature extraction.

Main Results:

• Standardized respiratory sounds from 84 patients, segmenting multiple respiratory cycles.
• Calculated average and standard deviation of amplitude features for each respiratory cycle segment.
• Observed distinct differences in respiratory sound features among various diseases.

Conclusions:

• A robust algorithm platform can segment respiratory sounds into inhale/exhale phases.
• The method allows for comparison of respiratory sound differences between diseases.
• Provides objective evidence for auscultation and visual display of breath sounds.