Brain scans of patients with multiple sclerosis will be evaluated by AI

Multiple sclerosis is a chronic disease that often leads to disability. More than 2.8 million people worldwide suffer from MS, most of them younger people. In addition to early diagnosis, treatment is particularly important, and today it can be tailored to each individual patient. Biomarkers from magnetic resonance imaging (MRI) of the brain and spinal cord play an important role in evaluating the effectiveness of treatment. MindGlide was developed to assess the effect of MS treatment on MRI biomarkers thanks to its perfect image recognition.

Fast and detailed

MindGlide is a publicly available deep learning model designed to quantify brain structures and lesions from any contrast-enhanced MRI image, detect changes in brain volume, structures, and lesions, and capture new lesions. It was trained on 4,247 images from 2,934 MS patients from 592 devices and validated on nearly 15,000 images from 1,001 patients from two clinical studies and a cohort of MS patients from routine care. 

Quantification of biomarkers from MR requires protocols with multiple contrast agents. However, most hospital archives contain images with a single contrast agent. MindGlide can extract brain regions and white matter lesions from MRI images with a single contrast agent. It can recognize how different types of treatment affect the progression of MS in clinical trials and in routine practice, both on images that were previously impossible to analyze and on routine images from routine practice. It only takes 5-10 seconds per image.

Artificial intelligence will thus soon end the era when measuring MRI biomarkers reflecting MS treatment outcomes requires different types of specialized MRI images, which many devices used in practice are unable to provide. Moreover, processing these images takes a very long time. 

Treasures from the archives

Thanks to AI, doctors will be able to determine and compare how different treatments affect the progression of MS, both in clinical studies and in routine care. The MindGlide model also outperformed two other artificial intelligence tools – SAMSEG (used to identify and outline different parts of the brain during MRI scans) and WMH-SynthSeg (detects and measures bright spots observed on certain MRI images). MindGlide was 60% better than SAMSEG and 20% better than WMH-SynthSeg at localising lesions or monitoring the effect of treatment.

As mentioned above, MindGlide is unique in that it enables quantitative analysis of single-contrast archival MRI scans. Study leader Dr. Philipp Goebl said that with the help of this tool, experts will be able to understand data from millions of unused archived brain images within a few years, which has been difficult or even impossible until now.

New goal: scanning the spinal cord

The current use of MindGlide is limited to brain images. It does not yet include the evaluation of spinal cord images, which play a crucial role in assessing disability in patients with MS. Researchers therefore face another goal – to develop a tool for evaluating the entire central nervous system.

Editorial Team, Medscope.pro

Sources: 

1. AI Tool can Track Effectiveness of Multiple Sclerosis Treatments. Digital Health News, 14. 4. 2025. Dostupné na www.digitalhealthnews.eu/research/7426-ai-tool-can-track-effectiveness-of-multiple-sclerosis-treatments
2. Goebl P., Wingrove J., Abdelmannan O. et. al.: Enabling new insights from old scans by repurposing clinical MRI archives for multiple sclerosis research. Nat Commun 2025 Apr 7; 16 (1): 3149, doi: 10.1038/s41467-025-58274-8.