Saved in:
Bibliographic Details
Main Authors: Shigeyasu, Yuki, Harada, Shota, Yoshizawa, Akihiko, Terada, Kazuhiro, Nakazima, Naoki, Kurata, Mariyo, Abe, Hiroyuki, Ushiku, Tetsuo, Bise, Ryoma
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.07539
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911100800335872
author Shigeyasu, Yuki
Harada, Shota
Yoshizawa, Akihiko
Terada, Kazuhiro
Nakazima, Naoki
Kurata, Mariyo
Abe, Hiroyuki
Ushiku, Tetsuo
Bise, Ryoma
author_facet Shigeyasu, Yuki
Harada, Shota
Yoshizawa, Akihiko
Terada, Kazuhiro
Nakazima, Naoki
Kurata, Mariyo
Abe, Hiroyuki
Ushiku, Tetsuo
Bise, Ryoma
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2508_07539
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Domain Generalization of Pathological Image Segmentation by Patch-Level and WSI-Level Contrastive Learning
Shigeyasu, Yuki
Harada, Shota
Yoshizawa, Akihiko
Terada, Kazuhiro
Nakazima, Naoki
Kurata, Mariyo
Abe, Hiroyuki
Ushiku, Tetsuo
Bise, Ryoma
Computer Vision and Pattern Recognition
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.
title Domain Generalization of Pathological Image Segmentation by Patch-Level and WSI-Level Contrastive Learning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2508.07539