Deep Learning-Based MRI Segmentation in MS: Human Study
In a study involving 57 patients with MS, the nnUNet architecture was compared to the DeepMedic model for segmenting lesions in 7T MRI scans. nnUNet achieved mean DSC values of 0.72, 0.71, and 0.66 for various MRI input types, with no significant differences noted compared to DeepMedic.
- Tags
- #MSResearch
- Most relevant to
- —
- Human efficacy
- Not established from this source
- Mechanism
- Not reported
- Evidence strength
- Not graded
- Outcome
- Not reported
Why it matters
These findings suggest that nnUNet performs comparably to existing models in lesion segmentation, which is crucial for improving MS diagnosis and monitoring.
What this does not prove
The research indicates the need for further studies to explore contrast-specific effects on small-lesion detection, particularly influenced by lesion size.
Next milestone
No next milestone was established from the available source.
Study facts
- Study design
- Not reported
- Participants / samples
- Not reported
- Randomised
- Not reported
- Controlled
- Not reported
- Primary endpoint met
- Not reported
- Relevant MS type
- Not reported
- Publication date
- 2026-10-14
- Evidence reviewed
- Abstract only
- Regulatory approval
- Not reported
- Research areas
- Not classified
- Biological target
- Not reported
- Pathway
- Not reported
- Population
- 57 patients with MS with longitudinal 7T MRI scans
- Size of effect
- mean DSC values of nnUNet: 0.72, 0.71, 0.66
- Secondary endpoints
- Not reported
- MRI endpoints
- Not reported
- Biomarkers
- Not reported
- Safety
- Not reported
Original sources
- Primary evidenceDeep Learning-Based Multiple Sclerosis Lesion Segmentation at 7 Tesla: Comparison of Single- and Multi-Contrast MRI Inputs. ↗DOI: 10.3390/jimaging12100510
Supporting passages (9)
publication datePublication date: 2026-10-14
interventionTitle: Deep Learning-Based Multiple Sclerosis Lesion Segmentation at 7 Tesla: Comparison of Single- and Multi-Contrast MRI Inputs.
comparatorDeepMedic was included as a comparative architecture.
findingsOn the test set, nnUNet achieved mean DSC values of 0.72, 0.71, and 0.66 for combined, MP2RAGE-only, and FLAIR-only inputs, respectively.
populationThe dataset included 115 longitudinal 7T MRI scans from 57 patients with MS, with 20 scans from 11 independent patients reserved for testing.
effect sizeOn the test set, nnUNet achieved mean DSC values of 0.72, 0.71, and 0.66 for combined, MP2RAGE-only, and FLAIR-only inputs, respectively.
limitationsTitle: Deep Learning-Based Multiple Sclerosis Lesion Segmentation at 7 Tesla: Comparison of Single- and Multi-Contrast MRI Inputs.
peer reviewedPublication types: Journal Article
subjectsThe dataset included 115 longitudinal 7T MRI scans from 57 patients with MS, with 20 scans from 11 independent patients reserved for testing.
AI assessment, not yet reviewed by a person · version 1 · Community votes are separate from evidence review.