<i>AdverIN:</i> Monotonic adversarial intensity attack for domain generalization in medical image segmentation

Zhang Z., Wang B., Yao L., Keles E., Jha D., Antalek M., ...More

MEDICAL IMAGE ANALYSIS, vol.107, 2026 (SCI-Expanded, Scopus) identifier identifier identifier

  • Publication Type: Article / Article
  • Volume: 107
  • Publication Date: 2026
  • Doi Number: 10.1016/j.media.2025.103848
  • Journal Name: MEDICAL IMAGE ANALYSIS
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, EMBASE, INSPEC, MEDLINE
  • Istanbul University Affiliated: No

Abstract

Domain generalization (DG) has emerged as a promising research direction because it can potentially enable deep learning models to handle data from previously unseen domains. DG methods try to achieve this by learning domain-invariant features that are robust to variations across different domains. This work proposes a novel domain generalization (DG) technique termed Adversarial Intensity Attack (AdverIN). AdverIN leverages an adversarial training strategy to augment data diversity by synthesizing a spectrum of intensity variations while preserving essential contextual information within images. We evaluated AdverIN through rigorous experiments on diverse multi-domain tasks: 2D retinal optic disc/cup segmentation and 3D prostate MRI segmentation. Our results demonstrate that AdverIN significantly improves the generalizability of segmentation models, achieving state-of-the-art performance on these challenging datasets. The code is available at https://github.com/NUBagciLab/AdverIN