analiza aiqbench
This research addresses a critical barrier to deploying AI respiratory classifiers in real-world settings by tackling the entanglement of stethoscope artifacts and actual disease signals. The proposed BTS-CAFE framework leverages causality-inspired interventions and counterfactual augmentation within a federated learning setup to create device-invariant decision boundaries, significantly improving out-of-distribution performance on unseen hardware. For practitioners in medical AI, this work offers a robust methodological path toward generalizable models that are not biased by specific recording equipment or demographic shortcuts.
Mitigating Stethoscope-Induced Shortcuts in Respiratory Sound Classification under Federated Domain Generalization with Causality-Inspired Interventions
arXiv:2605.29862v2 Announce Type: replace-cross Abstract: AI-driven respiratory sound classification (RSC) is promising for automated pulmonary disease detection, yet multi-site deployment is hindered by inter-stethoscope variability. We introduce a federated domain generalization (FedDG) formulation for RSC in which clients hold recordings from different stethoscopes and the model is evaluated on an unseen device. Our empirical analysis shows that stethoscope-induced style and disease-relevant content are partially entangled, making deterministic style removal unreliable. In response, we propose BTS-CAFE, a framework combining (i) causality-inspired device-style interventions with constraints designed to limit content distortion, (ii) counterfactual metadata augmentation to relieve device and demographic shortcuts, and (iii) gradient alignment to promote device-invariant decision boundaries across clients. Built on BTS with CLAP, a multimodal language-audio pretraining model, BTS-CAFE improves the out-of-distribution ICBHI Score by 3.69 points on average over five held-out devices relative to its backbone, and outperforms conventional data augmentation and federated learning baselines in simulated device leave-out evaluations on the ICBHI and SPRSound datasets.
Abstract:AI-driven respiratory sound classification (RSC) is promising for automated pulmonary disease detection, yet multi-site deployment is hindered by inter-stethoscope variability. We introduce a federated domain generalization (FedDG) formulation for RSC in which clients hold recordings from different stethoscopes and the model is evaluated on an unseen device. Our empirical analysis shows that stethoscope-induced style and disease-relevant content are partially entangled, making deterministic style removal unreliable. In response, we propose BTS-CAFE, a framework combining (i) causality-inspired device-style interventions with constraints designed to limit content distortion, (ii) counterfactual metadata augmentation to relieve device and demographic shortcuts, and (iii) gradient alignment to promote device-invariant decision boundaries across clients. Built on BTS with CLAP, a multimodal language-audio pretraining model, BTS-CAFE improves the out-of-distribution ICBHI Score by 3.69 points on average over five held-out devices relative to its backbone, and outperforms conventional data augmentation and federated learning baselines in simulated device leave-out evaluations on the ICBHI and SPRSound datasets.
Audio and Speech Processing (eess.AS); Artificial Intelligence (cs.AI); Sound (cs.SD)
arXiv:2605.29862 [eess.AS]
(or arXiv:2605.29862v2 [eess.AS] for this version)
https://doi.org/10.48550/arXiv.2605.29862 Focus to learn more arXiv-issued DOI via DataCite
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.
Text integral sub licența deschisă a publicației · 253 cuvinte
Subiecte conexe
[AINews] Opus 5.5 is good at explainer videos
Latent Space · 29 sept. 2026
Why do models task game?
AI Alignment Forum · 6 aug. 2026
RespFocus: Learning Temporal Evidence for Weakly Supervised Respiratory Sound Classification
arXiv · 15 oct. 2026
Biological Superintelligence
LessWrong · 30 iul. 2026