The Segment Anything Model (SAM) demonstrates impressive zero-shot segmentation ability on natural images but encounters difficulties in medical imaging due to domain shifts, anatomical variability, and its reliance on user-provided prompts. Recent prompt-free adaptations alleviate the need for expert intervention, yet still suffer from limited robustness and adaptability, often overlooking the issues of semantic over-smoothing and token uniformity. We propose SAM-DCE, which balances local discrimination and global semantics while mitigating token uniformity, enhancing inter-class separability, and enriching mask decoding with fine-grained, consistent representations. Extensive experiments on diverse medical benchmarks validate its effectiveness.
@article{arxiv.2509.16886,
title = {SAM-DCE: Addressing Token Uniformity and Semantic Over-Smoothing in Medical Segmentation},
author = {Yingzhen Hu and Yiheng Zhong and Ruobing Li and Yingxue Su and Jiabao An and Feilong Tang and Jionglong Su and Imran Razzak},
journal= {arXiv preprint arXiv:2509.16886},
year = {2025}
}