Particle.news
Download on the App Store
3 ARTICLES
·
last mo.

AI Reveals Thousands of Previously Hidden Cortical Lesions on Routine MS MRIs

The University at Buffalo team used a new multimodal enhancement pipeline to expose widespread gray‑matter damage on standard scans, with prospective clinical validation needed.

The neuroimaging framework engineered by Dr. Michael G. Dwyer and Dr. Robert Zivadinov at the University at Buffalo leverages the MMCLE generative AI algorithm to cross-reference legacy MRI image contrasts, exposing a massive hidden pathology of over 11,000 previously invisible gray matter cortical lesions across a multi-subject clinical trial dataset. Credit: Neuroscience News

Overview

  • A University at Buffalo‑led study published in Communications Medicine showed that AI can reveal cortical lesions that are not visible on individual conventional MRI images.
  • The researchers combined a new multimodal cortical lesion enhancement method called MMCLE with multi‑contrast deep learning to synthesize information across standard MRI sequences.
  • Applying the pipeline to legacy scans from the phase III ORATORIO trial of ocrelizumab, which contains more than 700 participants, the team identified roughly 15–20 cortical lesions per patient and over 11,000 lesions overall.
  • The authors say cortical lesions are closely linked to disability and cognitive decline in multiple sclerosis, so detecting them could change prognosis, explain persistent symptoms, and prompt reanalysis of past clinical trials.
  • The work is a large retrospective technical validation but not yet ready for routine care; the next steps are independent replication, prospective clinical studies, and development of workflows for use in research and clinics.