Using MRI-Machine Learning to Identify Disease Subtypes in Multiple Sclerosis: Alessia Bianchi, MD

WATCH TIME: 5 minutes | Captions are auto-generated and may contain errors.

"Our results support the application of machine learning (ML) in multiple sclerosis (MS), and this is a first big step. [The findings] demonstrated that ML can identify subtypes and stages of MS starting from MRI data. Also, [the study] supports the novel hypothesis of MS as a continuum, more than a distinct disease between relapsing and progressive."

Multiple sclerosis (MS) is a heterogeneous condition traditionally categorized by clinical features, with growing incorporation of MRI metrics into classification schemes. A recent study presented at the 2025 European Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS) Congress, held September 24-26, in Barcelona, Spain, demonstrated that machine learning (ML) applied to real-world MRI data could identify MS subtypes and stages, consistent with prior research.¹ Presented by lead author Alessia Bianchi, MD, the study cohort included MRI data from 306 patients with MS from 16 MAGNIMS centers and 104 healthy controls.

In the study, researchers used MRI scans, including T1-weighted and T2-FLAIR images, to train an unsupervised ML algorithm, SuStaIn, to identify disease subtypes and model temporal progression. Among 250 patients with complete MRI data from the study cohort, researchers identified 2 distinct MRI-driven subtypes. Findings showed that Subtype 1 (n = 235) was lesion-driven, demonstrating early increases in lesion volume and T1/FLAIR abnormalities, followed by atrophy of the cerebellum, corpus callosum, and spinal cord. For comparison, Subtype 2 (n = 15) displayed low lesion burden with early, diffuse cortical gray matter degeneration.

At ECTRIMS 2025, Bianchi, a postdoctoral research fellow at the University of Siena in Italy, provided background on the study in an interview with NeurologyLive^®. During the conversation, she noted that early-stage clustering of patients with clinically isolated syndrome and later-stage mapping of patients with progressive MS may support the emerging view of MS as a disease continuum. Additionally, she highlighted that disease stage correlated with both disease duration and Expanded Disability Status Scale scores, underscoring the potential of ML to refine MS classification and staging.

Click here for more coverage of ECTRIMS 2025.

REFERENCES
1. Bianchi A, Cortese R, Battaglini M, et al. Identifying MS Subtypes using MRI-Based Machine Learning. Presented at ECTRIMS Congress; September 24-26, 2025; Barcelona, Spain. Abstract P508.