aiqbench analysis
This paper introduces RespFocus, a weakly supervised framework utilizing multiple instance learning to extract interpretable temporal evidence from noisy respiratory recordings. By integrating cycle-level labels with diagnostic interventions, the method improves classification accuracy while addressing the "black box" nature of conventional deep learning models in clinical auscultation. The work advances machine learning interpretability specifically for biomedical signal processing applications.
RespFocus: Learning Temporal Evidence for Weakly Supervised Respiratory Sound Classification
arXiv:2610.15702v1 Announce Type: cross Abstract: Respiratory auscultation provides a low-cost means of assessing pulmonary function, yet transient crackles and sustained wheezes can be difficult to recognize in noisy clinical recordings. Conventional deep learning classifiers often operate as black boxes, offering limited insight into the temporal acoustic evidence supporting their predictions. We present RespFocus, a weakly supervised framework that uses multiple instance learning (MIL) to learn inspectable temporal evidence from cycle-level labels and integrate it into classification. We also introduce PedResp2026, a multicenter pediatric respiratory sound cohort collected in routine clinical environments with adjudicated cycle-level annotations. In classification experiments, RespFocus achieves four-class Scores of 69.55\% on public ICBHI 2017 and 72.91\% on PedResp2026. Evidence diagnostics and matched temporal interventions support the functional contribution of the learned evidence to model predictions. Together, these findings highlight RespFocus's potential as a tool for automated respiratory sound diagnosis.
Abstract:Respiratory auscultation provides a low-cost means of assessing pulmonary function, yet transient crackles and sustained wheezes can be difficult to recognize in noisy clinical recordings. Conventional deep learning classifiers often operate as black boxes, offering limited insight into the temporal acoustic evidence supporting their predictions. We present RespFocus, a weakly supervised framework that uses multiple instance learning (MIL) to learn inspectable temporal evidence from cycle-level labels and integrate it into classification. We also introduce PedResp2026, a multicenter pediatric respiratory sound cohort collected in routine clinical environments with adjudicated cycle-level annotations. In classification experiments, RespFocus achieves four-class Scores of 69.55\% on public ICBHI 2017 and 72.91\% on PedResp2026. Evidence diagnostics and matched temporal interventions support the functional contribution of the learned evidence to model predictions. Together, these findings highlight RespFocus's potential as a tool for automated respiratory sound diagnosis.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2610.15702 [cs.LG]
(or arXiv:2610.15702v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.15702 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 235 words
Related coverage
CoRE: Weakly Supervised Coarse-to-Fine Risk Evidence Learning in Driving Videos
arXiv · Aug 27, 2026
Behavioral Latency as Weak Event-Time Supervision for EEG Reaction-Time Decoding
arXiv · Sep 1, 2026
Unlocking Spatial Grounding in Large Audio-Visual Retrieval models
arXiv · Jul 29, 2026
Reinforcement Learning with Evolving Rubrics as Rewards for Audio Reasoning
arXiv · Aug 6, 2026