Fine-tuned U-Net-based model for automated WML segmentation on 7T MRI
Fine-tuning a U-Net-based model for automated segmentation of T2-hyperintense white matter lesions (WML) on 7T MRI improved performance significantly, achieving a median Dice score of 0.57 for T2-WML and notable results for T1-WML.
- Tags
- #Other
- Most relevant to
- —
Why it matters
This method enhances WML detection accuracy, which is crucial for multiple sclerosis research and could improve analysis in multi-center clinical settings.
What this does not prove
The findings indicate a drop in the original model's effectiveness when applied to 7T data compared to lower-field MRIs. Additionally, the study did not measure against other segmentation methods or provide demographic details of the data used.
Next milestone
No next milestone was established from the available source.
Study facts
- Study design
- Other
- Participants / samples
- Not reported
- Randomised
- Not reported
- Controlled
- Not reported
- Primary endpoint met
- Yes
- Relevant MS type
- Not reported
- Publication date
- 2026-09-08
- Evidence reviewed
- Abstract only
- Regulatory approval
- Not reported
- Research areas
- Other
Original sources
- Primary evidenceBridging field strengths: fine-tuned deep learning models for 7T MRI white matter lesion segmentation in multiple sclerosis. ↗DOI: 10.64898/2026.09.03.26362180
Supporting passages (10)
study phaseThis study evaluates fine-tuning as a domain adaptation strategy to leverage a pre-trained deep learning model for automated WML segmentation on 7T MRI.
study designFine-tuning is an effective and practical strategy for adapting deep learning WML segmentation algorithms to 7T MRI, even with limited annotated data.
subjectsusing a 7T dataset.
primary endpoint metFine-tuning markedly improved T2-WML segmentation, with the fine-tuned model achieving a median Dice score of 0.57.
research categoriesThe models presented here may be included in MS research workflows to facilitate multi-center collaborations with 7T MRI.
publication datePublication date: 2026-09-08
interventionWe fine-tuned a U-Net-based model, originally trained on approximately 35,000 heterogeneous lower-field (1T, 1.5T, 3T) multi-contrast MRI scans of people with MS for T2-hyperintense WML (T2-WML) segmentation, using a 7T dataset.
primary endpointMulti-center fine-tuning further improved performance, particularly for the more challenging task of T1-WML segmentation.
findingsFine-tuning markedly improved T2-WML segmentation, with the fine-tuned model achieving a median Dice score of 0.57.
limitationsTitle: Bridging field strengths: fine-tuned deep learning models for 7T MRI white matter lesion segmentation in multiple sclerosis.
AI assessment, not yet reviewed by a person · version 1 · Community votes are separate from evidence review.