Uncertainty‐guided test‐time optimization for personalizing segmentation models in longitudinal medical imaging
Uncertainty-guided TTO provides a robust, validation-free strategy for optimizing patient-specific segmentation models in longitudinal medical imaging. By tailoring personalization based on predictive uncertainty, our method improves segmentation quality across a range of imaging modalities and anatomical targets. This framework supports broad clinical deployment of personalized AI and motivates future extensions to contextual integration and multi-label segmentation. Code is publicly available at https://github.com/jchun-ai/uncertainty-tto.
Authors
- Matthew A. Firpo (ORCID: https://orcid.org/0000-0002-7983-3982)
- Mohamed Eltaher (ORCID: https://orcid.org/0009-0008-2078-4718)
- Kimberly S. Kirkwood (ORCID: https://orcid.org/0000-0002-6876-2752)
- Eugene J. Koay (ORCID: https://orcid.org/0000-0001-7675-3461)
- Aashish C. Gupta (ORCID: https://orcid.org/0000-0002-0178-0921)
- Caleb O'Connor
- Bilel Daoud
- Anirban Maitra (ORCID: https://orcid.org/0000-0001-7923-9978)
- McKell Woodland (ORCID: https://orcid.org/0000-0003-0995-3695)
- Jaehee Chun (ORCID: https://orcid.org/0000-0002-9695-6079)
- Mais M Al Taie
- Shanli Ding
- Austin H. Castelo (ORCID: https://orcid.org/0009-0009-3015-3570)
- Jeddy Bennett
- Kristy K. Brock
Institutions
- Brigham Young University (US)
- The University of Texas MD Anderson Cancer Center (US)
- University of California, San Francisco (US)
- University of Utah (US)
Publication Details
- Journal
- Medical Physics
- Published
- 2025-12-21
- DOI
- https://doi.org/10.1002/mp.70206
- Primary Topic
- Medical Image Segmentation Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Institutes of Health