Peer-reviewed article
Machine learning-based study of breath sounds in children with asthma in remission and cough variant asthma.
Hu C, Lv D, Liu J, Tang L, Chen Y et al. ·
Summary
Written by AI, not yet AICS's view · claude-haiku-4-5-20251001This study applied machine learning algorithms to analyze breath sounds in 50 children with asthma, comparing those in remission with those with cough variant asthma. Researchers used support vector machine, random forest, and logistic regression models to classify breath sound patterns. The support vector machine achieved 98.32% accuracy in distinguishing between the two asthma subtypes. The findings suggest machine learning could improve objective diagnosis of pediatric asthma.
Key points
- Support vector machine achieved 98.32% accuracy, 96.23% sensitivity, and 0.99 AUC in distinguishing asthma subtypes
- Study compared 30 children with asthma in remission and 20 with cough variant asthma
- Machine learning approach aims to reduce subjective variability in traditional physician auscultation
- Three algorithms tested: logistic regression, random forest, and support vector machine
Limitations
This is a small proof-of-concept study with only 50 children from a single setting. The high accuracy rates suggest potential overfitting and generalization to other populations and healthcare settings remains unclear.
Abstract
Background Traditional auscultation, heavily dependent on the subjective judgment of physicians, can lead to variability in diagnoses. This study aimed to explore the application of machine learning algorithms for analyzing breath sounds in children with asthma, particularly those with asthma in remission and those with cough variant asthma (CVA). Methods Our study collected breath sound data from 50 children with asthma (30 with asthma in remission and 20 with CVA). First, we preprocessed and extracted the breath sound data. Second, machine learning techniques were applied to objectively classify and evaluate the breath sounds of pediatric asthma patients. Then logistic regression, random forest, and support vector machine algorithms were employed to train models and predict outcomes. Results In this study, the support vector machine achieved the best performance in distinguishing between breath sounds from children with asthma in remission and those with CVA. It reached an accuracy of 98.32%, a sensitivity of 96.23%, and an area under the receiver operating characteristic curve of 0.99 in predicting pediatric asthma subtypes. Conclusions Our findings highlight the potential of machine learning models to revolutionize the diagnosis and treatment of pediatric asthma, offering a pathway towards more precise and individualized therapeutic strategies. Trial registration ClinicalTrials.gov number: ChiCTR2300077717 (Registration Date: 2023-11-16).
Findings the Hub has extracted
Each claim below is labelled by its type of design and links back to this report.
Problem
Subjective variability in diagnosis of pediatric asthma subtypes using traditional auscultation
- Intervention
- Machine learning algorithms (support vector machine, random forest, logistic regression) applied to analyze breath sounds for classification of asthma subtypes
- Population
- 50 children with asthma: 30 with asthma in remission and 20 with cough variant asthma
Key finding
Support vector machine achieved 98.32% accuracy in distinguishing breath sounds between asthma subtypes
C et al. · Europe PMC · 12 May 2026