Automated algorithm for Wet/Dry cough sounds classification.

Area of Science:

• Biomedical Engineering
• Respiratory Medicine
• Artificial Intelligence in Healthcare

Background:

• Cough is a prevalent respiratory symptom, crucial for differential diagnosis.
• Distinguishing between wet and dry coughs is clinically significant, often indicating bacterial infections.
• Current subjective methods limit long-term monitoring and treatment assessment.

Purpose of the Study:

• To develop automated technology for classifying coughs into wet and dry categories.
• To provide an objective tool for cough analysis beyond traditional clinical consultations.
• To enable long-term cough monitoring and assess treatment efficacy.

Main Methods:

• Development of novel features for cough sound analysis.
• Implementation of a Logistic regression-based model for wet/dry cough classification.
• Evaluation on a clinical database of pediatric and adult cough recordings using a non-contact microphone.

Main Results:

• The automated system achieved a sensitivity of 79±9% and specificity of 72.7±8.7% in classifying cough types.
• The proposed method demonstrated effectiveness on diverse patient populations (pediatric and adult).
• Successful classification of coughs into wet and dry categories using objective acoustic features.

Conclusions:

• The developed automated technology shows promise as a valuable clinical tool for cough monitoring.
• This method offers potential for objective, long-term assessment of respiratory conditions, particularly in home settings.
• Automated cough classification can supplement physician judgment and improve patient care for respiratory diseases.