通过局部与整体图像对比学习，提升病理图像分割在同院数据中的泛化能力。
Domain Generalization of Pathological Image Segmentation by Patch-Level and WSI-Level Contrastive Learning
- 利用非肿瘤区域特征聚类，将同医院不同样本视为不同域
- 设计两阶段对比学习，减少不同域间特征差异
- 适用于数据难跨院收集的病理图像分割场景
本文针对病理图像中的领域偏移问题，聚焦于同一医院内全切片图像（WSI）间的差异，如患者特征和组织厚度变化，而非跨医院差异。传统方法依赖多医院数据，但实际采集困难。为此，提出一种领域泛化方法：通过聚类非肿瘤区域的WSI级特征，将不同聚类结果视为不同域，并采用对比学习缩小不同域间特征差距。该方法引入两阶段对比学习策略——全切片级与局部块级对比学习，有效减小特征差异，提升模型在同院数据上的泛化性能。
原文摘要 · Abstract (English)
In this paper, we address domain shifts in pathological images by focusing on shifts within whole slide images~(WSIs), such as patient characteristics and tissue thickness, rather than shifts between hospitals. Traditional approaches rely on multi-hospital data, but data collection challenges often make this impractical. Therefore, the proposed domain generalization method captures and leverages intra-hospital domain shifts by clustering WSI-level features from non-tumor regions and treating these clusters as domains. To mitigate domain shift, we apply contrastive learning to reduce feature gaps between WSI pairs from different clusters. The proposed method introduces a two-stage contrastive learning approach WSI-level and patch-level contrastive learning to minimize these gaps effectively.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。