MPVT+: a noise-robust training framework for automatic liver tumor segmentation with noisy labels
The MPVT+ framework demonstrates that principled noise modeling, coupled with consistency training, effectively unlocks the potential of imperfect medical datasets. This strategy reduces the dependency on perfectly labeled datasets and moves fully automated liver tumor delineation closer to clinical applicability.
Authors
- Wei Lü (ORCID: https://orcid.org/0000-0001-6769-2695)
- Haoxiang Tian
- Tianshu Xie (ORCID: https://orcid.org/0000-0001-7021-8855)
- Jiajun Zhou (ORCID: https://orcid.org/0000-0002-5062-4183)
- Xuan Cheng
- Ming Liu (ORCID: https://orcid.org/0000-0002-4770-4449)
- Haigang Gong (ORCID: https://orcid.org/0000-0002-7176-4142)
- Yi Wei
Institutions
- University of Electronic Science and Technology of China (CN)
- Sichuan University (CN)
- Wenzhou Medical University (CN)
- West China Hospital of Sichuan University (CN)
- Quzhou University (CN)
Publication Details
- Journal
- Frontiers in Medicine
- Published
- 2025-09-10
- DOI
- https://doi.org/10.3389/fmed.2025.1653865
- Primary Topic
- AI in cancer detection
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- China Postdoctoral Science Foundation
- National Key Research and Development Program of China