Acoustic aware model improves respiratory sound classification accuracy
Clinically Informed Acoustic-Aware Attribute Modeling for Respiratory Sound Classification
Sound
Summary
Classifying lung sounds like crackles and wheezes helps doctors spot health issues. The authors noticed that these sounds overlap more like a mix of two attributes rather than separate categories. They created a new method that treats crackles and wheezes as distinct features and uses special timing techniques for each. This approach better matches how these sounds behave and scored higher on tests across different lung sound datasets.
What this means in practice
- •For medical device engineers: Develop lung sound analysis tools that detect crackles and wheezes more accurately by modeling their distinct acoustic patterns separately.
- •For hospital data teams: Integrate improved respiratory sound classification in patient monitoring systems for better detection of abnormal lung sounds.
- •For telemedicine service providers: Offer remote lung sound assessment services by deploying models that explicitly capture crackle and wheeze attributes for more reliable diagnoses.$Commercial implications: Supports advanced remote diagnostics tools appealing to telehealth providers and clinics seeking enhanced lung disease screening.
Authors
Xinhong Li, Li Xiao, Ziyang Chen, Yuhong Yang, Weiping Tu
Abstract
Automated respiratory sound classification (RSC) aims to identify normal and adventitious respiratory sounds, such as crackles and wheezes. Most existing methods formulate RSC as a flat four-class classification problem over norma, crackle, wheeze and both. However, unlike conventional classification tasks with semantically independent and mutually exclusive categories, the commonly used four-class RSC labels have a compositional structure and can be represented by two clinically meaningful sound attributes: crackle presence and wheeze presence. Existing flat classifiers do not explicitly exploit this structure, which may increase the learning burden by treating the four labels as independent targets. Meanwhile, crackles and wheezes exhibit distinct temporal characteristics, with sparse transient crackle events and sustained continuous wheeze patterns, making a single temporal aggregation strategy less suitable for modeling different adventitious sound evidence. To address these limitations, we propose an Acoustic-Aware Asymmetric Framework (AAAF) for RSC. AAAF reformulates four-class recognition as two sound-attribute detection tasks and recovers the final category through logical composition during inference. Built upon a pretrained BEATs encoder, AAAF employs branch-specific multi-layer fusion and asymmetric temporal aggregation to capture localized crackle evidence and sustained wheeze patterns. Auxiliary normal--abnormal supervision and a batch-adaptive dynamic loss are further introduced to improve global discrimination and robustness under imbalanced labels. Experiments on ICBHI and SPRSound demonstrate that AAAF improves overall RSC performance and generalizes across datasets, achieving a state-of-the-art four-class Score of 65.30% on ICBHI.