Abstract

Accurate liver MRI analysis is essential for liver disease assessment but re-mains challenging due to limited annotations. The CARE-Liver challenge ad-dresses this problem through two tasks: liver segmentation (LiSeg) and liver fibrosis staging (LiFS). For LiSeg, we propose an annotation-efficient segmen-tation framework based on pseudo-label generation and an adjacent-slice three-channel representation. For LiFS, we develop a mask-guided local patch framework with supervised contrastive learning for fibrosis stage classification. The proposed LiSeg framework achieves a Dice of 0.9727 and an HD of 3.62 mm in cross-validation. It further generalizes well to the official evaluation, achieving a Dice of 0.9623 on the validation set and Dice scores of 0.9715 and 0.9804 on the in-distribution (ID) and out-of-distribution (OOD) test cohorts, respectively. For LiFS, the proposed framework achieves a mean AUC of 0.8198 in cross-validation and maintains competitive performance on the offi-cial evaluation, with validation AUCs of 0.7632 and 0.7852 for Tasks 1 and 2, respectively, and test AUCs of 0.8163/0.8290 (Task 1) and 0.8363/0.5914 (Task 2) on the ID and OOD cohorts. Overall, these results demonstrate that the proposed framework effectively supports both liver segmentation and fi-brosis staging under limited annotation.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/CARE_034.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to Open Review

Open Review Page: https://openreview.net/forum?id=HptSxVlUee

BibTex

@InProceedings{HuaZij_PseudoLabelEnhanced_MICCAISAT2026,
        author = { Hua, Zijun AND Sun, Yujie AND Wu, Haoyu AND Sun, Xiaowu},
        title = { { Pseudo-Label-Enhanced Liver Segmentation and Mask-Guided Patch Learning for Fibrosis Staging Classification } },
        booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
        year = {2026},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 17273},
        month = {pending},
        page = {pending}
}


back to top