Machine Learning-Based Cough Tone Classification: Diagnostic Exploration of Chronic Obstructive Pulmonary Disease and...

Area of Science:

• Biomedical Engineering
• Computational Biology
• Pulmonology

Background:

• Distinguishing chronic obstructive pulmonary disease (COPD) from respiratory tract infections (RTI) is crucial for appropriate patient management.
• Current diagnostic methods can be invasive or time-consuming.
• Non-invasive, objective tools are needed for differential diagnosis.

Purpose of the Study:

• To develop and validate a non-invasive method for differentiating COPD from RTI using voice signal analysis.
• To apply machine learning models to voice features for classification.
• To assess the performance of Logistic Regression (LR) and Random Forest (RF) models.

Main Methods:

• Collected fixed-pattern voice signals from 25 COPD patients and 25 RTI patients.
• Analyzed over 400 multi-dimensional voice features to identify significant differences.
• Selected 18 statistically significant features (P < 0.05) and applied dimensionality reduction.
• Trained and evaluated LR and RF models for classification.

Main Results:

• Eighteen voice features significantly differentiated COPD from RTI patients.
• The Logistic Regression (LR) model achieved an Area Under the Curve (AUC) of 0.95.
• The Random Forest (RF) model achieved an AUC of 0.76, performing significantly lower than LR.

Conclusions:

• Voice analysis combined with machine learning, particularly the LR model, is a feasible and promising non-invasive tool for differentiating COPD from RTI.
• This approach provides a foundation for clinical application in differential diagnosis of respiratory conditions.
• Further optimization of this voice-based diagnostic method is warranted.