This challenge tackles multi-label classification for known chest X-ray (CXR) lesions and zero-shot classification for unseen ones. To handle diverse CXR projections, we integrate projection-specific models via a classification network into a unified framework. For zero-shot classification (Task 2), we extend CheXzero with a novel dual-branch architecture that combines contrastive learning, Asymmetric Loss (ASL), and LLM-generated descriptive prompts. This effectively mitigates severe long-tail imbalances and maximizes zero-shot generalization. Additionally, strong data and test-time augmentations (TTA) ensure robustness across both tasks.
@article{arxiv.2604.02185,
title = {CXR-LT 2026 Challenge: Projection-Aware Multi-Label and Zero-Shot Chest X-Ray Classification},
author = {Juno Cho and Dohui Kim and Mingeon Kim and Hyunseo Jang and Chang Sun Lee and Jong Chul Ye},
journal= {arXiv preprint arXiv:2604.02185},
year = {2026}
}
Comments
5 pages, 3 figures. Accepted to the IEEE ISBI 2026 CXR-LT Challenge