University at Buffalo researchers have used artificial intelligence to reveal previously invisible brain lesions in multiple sclerosis patients by analyzing existing MRI scans. The method, which uncovered more than 11,000 cortical lesions in a dataset of over 700 people, provides a new window into gray matter damage that has long been linked to cognitive decline and disability but remained undetectable on conventional imaging.
The work, published in Communications Medicine, addresses a decades-old frustration in MS care and research: gray matter lesions are recognized as a key driver of disease progression, yet standard MRI cannot see them. "We have all been very frustrated, knowing that these cortical lesions were there but not being able to see them," said Michael G. Dwyer, PhD, first author and associate professor of neurology and biomedical informatics at the Jacobs School of Medicine and Biomedical Sciences at UB.
Invisible damage becomes visible
Cortical lesions have been documented in postmortem tissue for more than a century, but clinical MRI resolves only white matter lesions. Even when diagnostic criteria began including cortical lesions in the 21st century, their usefulness remained limited by imaging technology. The UB-led team changed that by applying generative AI to existing scans from the phase III ORATORIO clinical trial of ocrelizumab.
Dwyer explained that individual images largely showed white matter damage. When multiple contrast images were processed together, however, the AI detected subtle tissue differences. "Because it sees those minor discrepancies, AI can reveal that there's something going wrong there, that the tissue is not behaving like healthy tissue," Dwyer said. The approach, called multimodal cortical lesion enhancement (MMCLE), synthesized what had been missing from single-scan readings. This application of AI for Healthcare demonstrates how computational methods can extract clinically meaningful data from imaging that already exists in patient records.
Thousands of hidden lesions found
The researchers applied the technique to MRI data from 732 participants in the ORATORIO trial, which tested ocrelizumab against a placebo. The AI revealed between 15 and 20 cortical lesions per patient-more than 11,000 lesions across the full dataset-that were invisible on the original scans. "If you look on the original scans, you generally can't see the cortical lesions," Dwyer said.
Robert Zivadinov, MD, PhD, senior author and director of the Buffalo Neuroimaging Analysis Center, said the findings will reshape both historical data reviews and future research. "This work, which has revealed that there is so much invisible pathology in the brain, will have tremendous impact for reviewing data from past clinical trials and also for those going forward," he said. Access to such indicators without new scanning hardware opens possibilities for retroactive study and real-world monitoring-pushing forward AI for Science & Research in neurology.
Why this matters for Science and Research
For researchers working with clinical trial data or longitudinal patient records, the ability to detect cortical lesions retrospectively changes what can be learned from past studies. It provides a direct, quantitative measure of disease progression in the gray matter that was previously only accessible through autopsy. The method does not require new scanner investment-only the application of AI to already-collected images-which could accelerate the validation of new MS therapies and the selection of trial endpoints that better reflect cognitive outcomes. Teams designing imaging protocols or reanalyzing completed trials can now incorporate cortical lesion load as a meaningful variable without altering acquisition sequences.
Your membership also unlocks:
