A classification framework for identifying bronchitis and pneumonia in children based on a small-scale cough sounds...

Area of Science:

• Pediatric respiratory medicine
• Medical acoustics
• Machine learning in healthcare

Background:

• Pneumonia is a leading cause of mortality in children globally, necessitating improved diagnostic tools.
• Bronchitis and pneumonia are common pediatric respiratory conditions that place significant burdens on healthcare systems.

Purpose of the Study:

• To develop and evaluate a novel framework for classifying pediatric bronchitis and pneumonia using cough sound analysis.
• To investigate the efficacy of machine learning models in differentiating these respiratory diseases based on acoustic features.

Main Methods:

• A Classification Framework based on Cough Sounds (CFCS) was proposed, utilizing cough sound recordings from 173 pediatric outpatients.
• Aggregation operations were employed to extract relevant disease features from cough sound segments.
• Support Vector Machine (SVM) was used for initial classification on the raw dataset, followed by data augmentation and Long Short-Term Memory Network (LSTM) for enhanced classification.

Main Results:

• The SVM model achieved a classification accuracy of 86.04% on the raw dataset, with high precision and recall for both bronchitis and pneumonia.
• Data augmentation, specifically pitch-shifting, improved model performance, with SVM and LSTM models reaching AUC values of 0.92 and 0.93, respectively.
• The CFCS demonstrated effectiveness in distinguishing between bronchitis and pneumonia in pediatric patients.

Conclusions:

• The Classification Framework based on Cough Sounds (CFCS) offers a promising, non-invasive method for diagnosing pediatric respiratory infections.
• Cough sound analysis, combined with machine learning, can effectively aid in the differential diagnosis of bronchitis and pneumonia in children.
• Further research and validation of the CFCS could lead to improved clinical decision-making and patient outcomes.