analiza aiqbench
This external validation study tests whether saliency maps can flag errors in chest radiograph classifiers at hospitals other than the one where the model was trained. Grad-CAM localized findings less often than radiologists, and adding saliency to a confidence-based error detector gave at most a 0.011 gain in AUC, with the signal reversing between sites for one model. For teams deploying medical imaging AI, the practical takeaway is that model confidence is a more reliable monitoring signal than saliency, and any explanation method needs external validation before anyone relies on it as a safeguard.
Can saliency maps flag chest radiograph classifier errors at a new hospital? An external validation study
arXiv:2611.05011v1 Announce Type: cross Abstract: Saliency maps are offered as a safeguard for medical image classifiers, but whether they reveal individual errors at a new hospital remains largely untested. We trained a convolutional network and a vision transformer on NIH ChestX-ray14 and evaluated 303,768 decisions per model at the training hospital and at two external sites, in the USA and Vietnam, with radiologist-consensus labels. We scored each explanation by its anatomical enrichment, the area-normalized concentration of attribution in a segmented thoracic region, and tested whether it detects errors beyond the model's confidence. Grad-CAM localized findings less often than radiologists (hit rate at most 0.575 versus 0.888). Adding enrichment to a confidence-based error detector calibrated at the training hospital never raised its area under the curve by more than 0.011, and for the convolutional network the relation between enrichment and correctness reversed between hospitals, degrading the transferred detector. Model confidence, not saliency, should anchor deployment monitoring.
Title:Can saliency maps flag chest radiograph classifier errors at a new hospital? An external validation study
Abstract:Saliency maps are offered as a safeguard for medical image classifiers, but whether they reveal individual errors at a new hospital remains largely untested. We trained a convolutional network and a vision transformer on NIH ChestX-ray14 and evaluated 303,768 decisions per model at the training hospital and at two external sites, in the USA and Vietnam, with radiologist-consensus labels. We scored each explanation by its anatomical enrichment, the area-normalized concentration of attribution in a segmented thoracic region, and tested whether it detects errors beyond the model's confidence. Grad-CAM localized findings less often than radiologists (hit rate at most 0.575 versus 0.888). Adding enrichment to a confidence-based error detector calibrated at the training hospital never raised its area under the curve by more than 0.011, and for the convolutional network the relation between enrichment and correctness reversed between hospitals, degrading the transferred detector. Model confidence, not saliency, should anchor deployment monitoring.
21 pages, 5 figures, 1 table, 6 Extended Data figures and 4 Extended Data tables. Code: this https URL
Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Image and Video Processing (eess.IV)
arXiv:2611.05011 [cs.CV]
(or arXiv:2611.05011v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2611.05011 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.
Text integral sub licența deschisă a publicației · 290 cuvinte
Subiecte conexe
Evaluation Choices Shape Biomedical ML Claims: A Pediatric Pneumonia Benchmark Case Study
arXiv · 30 sept. 2026
Bayesian uncertainty estimation improves clinical decision making in medical AI agents
arXiv · 24 iul. 2026
Reliability- and Anatomy-Consistency-Aware Multimodal Learning for Robust Fracture Classification from Bangladeshi Radiographs
arXiv · 25 aug. 2026
Knee-xRAI: An Explainable AI Framework for Automatic Kellgren-Lawrence Grading of Knee Osteoarthritis
arXiv · 9 iun. 2026