AI-Based Model Reframes Multiple Sclerosis as a Continuous Disease with Dynamic Stages

A groundbreaking international study, published on August 20, 2025, in Nature Medicine, has challenged the traditional view of multiple sclerosis (MS) as a disease with fixed subtypes like relapsing or progressive forms. Led by researchers from the Medical Center at the University of Freiburg and the University of Oxford, the study analyzed data from over 8,000 patients and more than 35,000 MRI scans, sourced from the NO.MS cohort (sponsored by Novartis), Roche Ocrelizumab cohort, and MS PATHS cohort.

The researchers propose a paradigm shift: instead of categorizing MS into rigid subtypes, the disease should be understood as a spectrum characterized by continuous pathological processes. Using artificial intelligence, the team identified four core disease dimensions: physical disability, brain damage, clinical relapses, and silent inflammatory activity. This approach allows for a more nuanced understanding of disease progression, recognizing MS as a dynamic system with specific state transitions rather than fixed categories.

The probabilistic model developed describes the progression of MS as a sequence of states with defined transition probabilities. Early, less severe stages typically involve inflammatory activity, gradually advancing to irreversible, severe stages. Importantly, the study indicates that direct progression into severe disease stages without prior inflammatory activity is virtually impossible, emphasizing the role of silent inflammation or relapses as key drivers of deterioration.

One of the significant implications of this research is the potential to transform diagnostics and treatment protocols. Current approvals are often based on predefined subtypes, which may limit access to effective therapies. The new model enables personalized risk assessments and dynamic monitoring, allowing clinicians to tailor interventions based on a patient's current disease state rather than static labels.

Prof. Dr. Heinz Wiendl from the University of Freiburg highlights that "MS is now understood as a continuous disease process with observable state transitions." This perspective promotes early, individualized treatment, especially for patients with silent inflammatory activity that could be overlooked in traditional categorization.

Beyond MS, the principles of this data-driven, state-based modeling have broad applications. Experts such as Prof. Dr. Lutz Hein and Prof. Dr. Peter Berlit emphasize that such models can extend to various neurological and other diseases, facilitating more flexible and precise disease management. The model has been validated using external datasets, paving the way for integration into clinical practice, treatment decision-making, and patient education.

Overall, this innovative approach signifies a major advancement in understanding MS and offers a template for future disease classification systems that prioritize patient-specific disease dynamics over rigid categories.