A pathology-attention multi-instance learning framework for multimodal classification of colorectal lesions
These results demonstrate that PAT-MIL effectively mitigates staining variability and enhances cross-center generalization through the collaborative modeling of visual and textual modalities. It provides a robust solution for colorectal lesion classification without relying on pixel-level annotations, advancing the field of multimodal pathological image analysis.
Authors
- Mingye Fu (ORCID: https://orcid.org/0000-0002-4692-1062)
- Jin‐Shun Lin
- Lianghui Zhu (ORCID: https://orcid.org/0009-0005-3067-0210)
- Lin Xie (ORCID: https://orcid.org/0000-0002-8850-6764)
- Wenbin Dai (ORCID: https://orcid.org/0000-0001-5113-3952)
- Jianfeng Wu (ORCID: https://orcid.org/0000-0002-9501-7096)
- Tian Guan (ORCID: https://orcid.org/0000-0003-2509-2952)
- Freddie H. Fu (ORCID: https://orcid.org/0000-0001-7821-4269)
- Xuemei Zhang (ORCID: https://orcid.org/0000-0002-4979-6965)
- Zhe Wang (ORCID: https://orcid.org/0000-0002-4338-2630)
- Jing Peng (ORCID: https://orcid.org/0000-0002-3305-8773)
- Yonghong He (ORCID: https://orcid.org/0000-0002-9964-8901)
- Zhaoxuan Wang (ORCID: https://orcid.org/0000-0001-8169-7978)
Institutions
- Washington University in St. Louis (US)
- University Town of Shenzhen (CN)
- Tsinghua–Berkeley Shenzhen Institute (CN)
- Fourth Affiliated Hospital of Guangxi Medical University (CN)
- Shenzhen University Health Science Center (CN)
- Xijing Hospital (CN)
- Air Force Medical University (CN)
Publication Details
- Journal
- Frontiers in Pharmacology
- Published
- 2025-06-06
- DOI
- https://doi.org/10.3389/fphar.2025.1592950
- Citations
- 1
- Primary Topic
- Colorectal Cancer Screening and Detection
- Type
- article
- Field-Weighted Citation Impact
- 1.05
Funders
- National Key Clinical Specialty Discipline Construction Program of China