Multiple sclerosis cortical lesion detection with deep learning at ultra-high-field MRI
Why is this work in the frame?
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
The three-model screen
all 1,000 screened works →All three models called this out of scope.
Deep learning framework for detecting MS cortical lesions; a clinical imaging method, and the reliability language is domain measurement, not the reproducibility literature.
The work develops a deep-learning tool for detecting multiple-sclerosis lesions.
Deep-learning tool for detecting MS cortical lesions is clinical imaging methods development, not study of research methods.
Abstract
Abstract Manually segmenting multiple sclerosis (MS) cortical lesions (CL) is extremely time-consuming, and past studies have shown only moderate inter-rater reliability. To accelerate this task, we developed a deep learning-based framework (CLAIMS: Cortical Lesion Artificial Intelligence-based assessment in Multiple Sclerosis) for the automated detection and classification of MS CL with 7T MRI. Two 7T datasets, acquired at different sites, were considered. The first consisted of 60 scans that include 0.5mm isotropic MP2RAGE acquired 4 times (MP2RAGEx4), 0.7mm MP2RAGE, 0.5mm T2*-weighted GRE, and 0.5mm T2*-weighted EPI. The second dataset consisted of 20 scans including only 0.75×0.75×0.9 mm MP2RAGE. CLAIMS was first evaluated using 6-fold cross-validation with single and multi-contrast 0.5mm MRI input. Second, performance of the model was tested on 0.7mm MP2RAGE images after training with either 0.5mm MP2RAGEx4, 0.7mm MP2RAGE, or alternating the two. Third, its generalizability was evaluated on the second external dataset and compared with a state-of-the-art technique based on partial volume estimation and topological constraints (MSLAST). CLAIMS trained only with MP2RAGEx4 achieved comparable results to the multi-contrast model, reaching a CL true positive rate of 74% with a false positive rate of 30%. Detection rate was excellent for leukocortical and subpial lesions (83%, and 70%, respectively), whereas it reached 53% for intracortical lesions. The correlation between disability measures and CL count was similar for manual and CLAIMS lesion counts. Applying a domain-scanner adaptation approach and testing CLAIMS on the second dataset, the performance was superior to MSLAST when considering a minimum lesion volume of 6μL (lesion-wise detection rate of 71% vs 48%). The proposed framework outperforms previous state-of-the-art methods for automated CL detection across scanners and protocols. In the future, CLAIMS may be useful to support clinical decisions at 7T MRI, especially in the field of diagnosis and differential diagnosis of multiple sclerosis patients.
Stored with the screening record, where it is evidence for the labels above.
The record
- Venue
- medRxiv
- Topic
- Multiple Sclerosis Research Studies
- Field
- Medicine
- Canadian institutions
- Université du Québec à Trois-RivièresMcGill UniversityMontreal Neurological Institute and Hospital
- Funders
- National Institute of Neurological Disorders and StrokeNational Institutes of HealthCentre Hospitalier Universitaire VaudoisCentre d'Imagerie BioMédicaleNational Institute of Mental HealthUniversité de LausanneUniversité de GenèveHôpitaux Universitaires de GenèveÉcole Polytechnique Fédérale de LausanneSanofiEuropean CommissionMultiple Sclerosis SocietyUniversitätsspital BaselNational Multiple Sclerosis Society
- Keywords
- Generalizability theoryMultiple sclerosisLesionMedicineNuclear medicineArtificial intelligenceComputer sciencePathologyPsychology
- Has abstract in OpenAlex
- yes