Useful nonrobust features are ubiquitous in biomedical images

arXiv:2604.22579v1 Announce Type: cross Abstract: We study whether deep networks for medical imaging learn useful nonrobust features - predictive input patterns that are not human interpretable and highly susceptible to small adversarial perturbations - and how these features impact test performance. We show that models trained only on nonrobust features achieve well above chance accuracy across five MedMNIST classification tasks, confirming their predictive value in-distribution. Conversely, adversarially trained models that primarily rely on robust features sacrifice in-distribution accuracy but yield markedly better performance under controlled distribution shifts (MedMNIST-C). Overall, nonrobust features boost standard accuracy yet degrade out-of-distribution performance, revealing a practical robustness-accuracy trade-off in medical imaging classification tasks that should be tailored to the requirements of the deployment setting.

Related

• M-IDoL: Information Decomposition for Modality-Specific and Diverse Representation Learning in Medical Foundation Model
• Beyond Attack Success Rate: A Multi-Metric Evaluation of Adversarial Transferability in Medical Imaging Models
• Rethinking Cross-Dose PET Denoising: Mitigating Averaging Effects via Residual Noise Learning
• Are Natural-Domain Foundation Models Effective for Accelerated Cardiac MRI Reconstruction?

Source: arXiv cs.CV | 2026-04-27