AI Model Distinguishes EDSS Subgroup Patterns in Multiple Sclerosis, Study Shows

A large-scale analysis published in the Multiple Sclerosis Journal examined patients with multiple sclerosis (MS) who shared identical Neurostatus-Expanded Disability Status Scale (EDSS) scores and revealed disability patterns distinguishable through artificial intelligence (AI) based clustering algorithms.

Led by Martina Greslin, PhD candidate and neurologist in the Department of Neurology at the University Hospital Basel in Switzerland, results from the analysis indicated that AI clustering yielded grouping assessments with similar patterns, including patients with identical EDSS scores of at least 4 who were further grouped into 4 subscore patterns. After feature exclusion, the data set was composed of 15 subscores derived from three FS: Pyramidal, Cerebellar and Sensory.

“We applied machine learning algorithms (MLA) on those subscores and identified clusters which led to distinct disability patterns within EDSS scores ranging from 4.0 to 6.5,” the study authors wrote. “These patterns allow describing with more granularity the range of functional impairment beyond ambulation in pwMS. In our data set, higher EDSS scores include a higher proportion of assessments falling into pattern A. This might indicate that these patterns may also be useful for a more granular quantification of the severity of deficits or further prognosis, but further studies are necessary to explore this option.”

Greslin et al continued, “such increased granularity may increase the power of clinical trials to detect treatment effects on deficits that are relevant to patients and their ADL but are not reflected by quantifying ambulation only. A more comprehensive description of the burden of disease may help to better select patients with MS who may benefit from more specific treatments, both symptomatic and targeted according to underlying pathogenic processes.”

The analysis was comprised of 13,103 assessments from 1636 patients with secondary progressive MS who participated in the double-blind, randomized, phase 3 EXPAND (NCT01665144) trial. The dataset included Functional System scores (FSS) and their corresponding subscores, Ambulation scores (AS), and EDSS scores. A descriptive analysis was conducted to identify relevant function systems with the subscores then binarized based on the Neurostatus definition and grouped by respective EDSS scores. After that, two consecutive MLAs were used to cluster the data, and new subscore patterns were created by aggregating clusters based on their dominant features.

In the analysis, the 4 distinct subscore patterns were identified within each EDSS score ranging by 0.5 increments from 4 to 6.5. Pattern A was defined as a cluster in which patients had high impact on more than 50% of activities of daily living (ADL) assessments, including impairments in muscle strength, tremor/dysmetria, truncal ataxia, or gait ataxia. Pattern B was characterized by high scores in spasticity or tandem walking whereas pattern C was defined by high values in only one of the sensory features or the Romberg test. Pattern D included assessments in which none of these features were elevated.

READ MORE: NeurologyLive® Year in Review 2025: Top Interviews on the AI Takeover

The continued use of AI in the field of neurology has become more commonplace within the last few years. Advances in machine learning, imaging analysis, and data integration have supported new approaches to diagnosis, risk stratification, and disease monitoring across neurologic conditions.

In early 2025, in collaboration with Cleveland Clinic, NeurologyLive® held a roundtable discussion with MS experts who talked about key advancements in MS treatment and research that were discussed during the 2025 Americas Committee for Treatment and Research in Multiple Sclerosis (ACTRIMS) Forum. The panelists, including neuroimmunologist Moein Amin, MD, neurologist Marisa McGinley, DO, and neurologist Devon Conway, MD, explored how artificial intelligence and machine learning can be potentially integrated into MS research and practice.

REFERENCE
1. Greselin M, Lu P-J, Mroczek M, et al. AI-assisted identification of disability patterns within identical EDSS grades. Multiple Sclerosis Journal. 2025;31(6):677-688. doi:10.1177/13524585251327300