arXiv：cs.LG（机器学习，全量分类）· Ahmed Sharshar, Asif Hanif, Naveen Kumar Kummari, Mohammad Yaqub, Mohsen Guizan·· 15 小时前AI 评分37
PhaseAT：面向医学图像域泛化的傅里叶相位对抗训练
PhaseAT: Fourier Phase Adversarial Training for Medical Image Domain Generalization
AI 导读
PhaseAT 是一种相位感知对抗训练框架，通过在傅里叶域迭代更新有界相位扰动、保持幅度谱不变来构建训练视图，并仅作用于 YCbCr 亮度通道，配合相位显著性掩码聚焦关键频率。该方法在单源域泛化上取得超过 20% 的提升，优于多种 SOTA DG 方法，论文已被 MICCAI 2026 接收，代码已开源。
正文
Abstract:Reliable clinical deployment of deep medical image models is hindered by distribution shifts across scanners, sites, and acquisition protocols. Existing domain generalization (DG) methods often focus on style or intensity diversification, but they can still leave networks dependent on domain-specific texture correlations. Inspired by evidence that Fourier phase encodes semantic structure, we introduce PhaseAT, a phase-aware adversarial training framework for medical DG. PhaseAT forms phase-perturbed training views in the Fourier domain by iteratively updating a bounded phase perturbation while keeping the amplitude spectrum unchanged, thereby stressing spatial organization under matched appearance statistics. Perturbations are applied only to the luminance channel in YCbCr color space to avoid chromatic artifacts. Additionally, a simple phase-saliency mask concentrates updates on the most influential frequencies. The model is trained with a weighted combination of losses on clean and phase-perturbed samples, supporting both single-source and multi-source DG. We validate our method on two challenging medical datasets and demonstrate that PhaseAT achieves over 20% improvement in single-source domain generalization, outperforming several state-of-the-art DG methods. The code implementation is available at: this https URL.
| Comments: | The paper is accepted in MICCAI 2026 |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.01807 [cs.CV] |
| (or arXiv:2610.01807v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.01807 arXiv-issued DOI via DataCite (pending registration) |
|
| Journal reference: | Medical Image Computing and Computer Assisted Intervention - MICCAI 2026, Lecture Notes in Computer Science, vol. 16881, pp. 413-423, Springer, 2027 |
| Related DOI: | https://doi.org/10.1007/978-3-032-38072-2_40
DOI(s) linking to related resources |
Submission history
From: Ahmed Sharshar [view email]
[v1]
Thu, 1 Oct 2026 14:46:32 UTC (746 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org