Identifying multiple sclerosis subtypes using unsupervised machine learning and MRI data

Arman Eshaghi,Charles R G Guttmann,Douglas L Arnold,Sridar Narayanan,Peter A Wijeratne,Declan Chard,Alan J Thompson,Alexandra L Young,Daniel C Alexander,Olga Ciccarelli,Ferran Prados,Frederik Barkhof

doi:10.1038/s41467-021-22265-2

Abstract

Multiple sclerosis (MS) can be divided into four phenotypes based on clinical evolution. The pathophysiological boundaries of these phenotypes are unclear, limiting treatment stratification. Machine learning can identify groups with similar features using multidimensional data. Here, to classify MS subtypes based on pathological features, we apply unsupervised machine learning to brain MRI scans acquired in previously published studies. We use a training dataset from 6322 MS patients to define MRI-based subtypes and an independent cohort of 3068 patients for validation. Based on the earliest abnormalities, we define MS subtypes as cortex-led, normal-appearing white matter-led, and lesion-led. People with the lesion-led subtype have the highest risk of confirmed disability progression (CDP) and the highest relapse rate. People with the lesion-led MS subtype show positive treatment response in selected clinical trials. Our findings suggest that MRI-based subtypes predict MS disability progression and response to treatment and may be used to define groups of patients in interventional trials.

Highlights

Multiple sclerosis (MS) can be divided into four phenotypes based on clinical evolution
We have recently developed an unsupervised machine learning algorithm, called Subtype and Staging Inference (SuStaIn)[6], to uncover data-driven disease subtypes with distinct temporal progression patterns
Of the 18 magnetic resonance imaging (MRI) features measured, 13 significantly differed between the MS training dataset and control group, and these were retained in the SuStaIn model

Summary

Introduction

Multiple sclerosis (MS) can be divided into four phenotypes based on clinical evolution. The consequences of redefining disease subtypes based on biology rather than on clinical grounds alone are that clinical trials should be better able to recruit patients who are likely to benefit from the medication under investigation New technologies, such as artificial intelligence and machine learning, can evaluate multidimensional data to identify groups with similar features. Such methods, applied to visible abnormalities on MRI scans, have great promise in classifying patients who share similar pathobiological mechanisms rather than common clinical features[1]. Current practice divides MS into four phenotypes: clinically isolated syndrome (CIS), relapsing-remitting MS (RRMS), primaryprogressive MS (PPMS) and secondary progressive MS (SPMS)[3] Two descriptors underly these phenotypes: (i) disease activity, as evidenced by relapses or new activity on magnetic resonance imaging (MRI) and (ii) progression of disability[3]. Once the SuStaIn subtypes and their MRI trajectories are identified, the resulting disease model can determine how closely a patient, whose MRI is unseen, belongs to each subtype and stage

Objectives

Methods

Results

Conclusion