Repurposing Clinical MRI Archives for Multiple Sclerosis Research with a Flexible, Single-Modality Approach: New Insights from Old Scans
Bibliographic record
Abstract
ABSTRACT In multiple sclerosis (MS), magnetic resonance imaging (MRI) biomarkers are critical for research in diagnosis, prognosis and assessing treatment efficacy. Traditionally, extracting relevant biomarkers of disease activity and neurodegeneration requires multimodal MRI protocols, limiting the use of the already existing vast amount of incomplete or single-modality MRI data which are acquired in clinical settings. We developed MindGlide, a deep learning model that extracts volums of brain regions and lesion from a single MRI modality, simplifying analysis and enabling the use of heterogeneous clinical archives. We trained MindGlide on a dataset of 4,247 brain MRI scans from 2,934 MS patients across 592 MRI scanners and validated it on 14,952 brain MRI scans from 1001 patients from three unseen external validation cohorts including 161 adolescent patients. Using dice scores, we demonstrated that MindGlide accurately estimated white matter lesion, cortical, and deep grey matter volumes. These volumes correlated with disability (Expanded Disability Status Scale, absolute correlation coefficients 0.1-0.2, p<0.05), and MindGlide outperformed an established tool in this regard. MindGlide robustly detected treatment effects across clinical trials, including disease activity and neurodegeneration (as shown by lesion accrual and brain tissue loss, respectively), even when analysing MRI modalities not traditionally used for such detailed measurements. Our results indicate the potential to indirectly reduce scan time and drug development costs in clinical trials while directly transforming the utility of retrospective analysis of real-world data acquired in clinical settings. As a consequence, scan time will be reduced and, in turn, the cost of trials.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.006 | 0.012 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".