Domain Generalization of Pathological Image Segmentation by Patch-Level and WSI-Level Contrastive Learning
This addresses domain generalization in medical imaging for pathologists, but it is incremental as it builds on existing contrastive learning techniques.
The paper tackles domain shifts within whole slide images (WSIs) for pathological image segmentation by proposing a two-stage contrastive learning method that clusters WSI-level features from non-tumor regions and applies patch-level and WSI-level contrastive learning to reduce feature gaps, achieving improved generalization without multi-hospital data.
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.