Defining multiple sclerosis subtypes using machine learning

Abstract

Multiple sclerosis (MS) is subdivided into four phenotypes on the basis of medical history and clinical symptoms. These phenotypes are defined retrospectively and lack clear pathobiological underpinning. Since Magnetic Resonance Imaging (MRI) better reflects disease pathology than clinical symptoms, we aimed to explore MRI-driven subtypes of MS based on pathological changes visible on MRI using unsupervised machine learning. In separate train and external validation sets we looked at a total of 21,170 patient-years of data from 15 randomised controlled trials and three observational cohorts to explore MRI-driven subtypes and test whether these subtypes had differential clinical outcomes. We processed MRI data to obtain measures of brain volumes, lesion volumes, and normal appearing white matter T1/T2. We identified three MRI-driven subtypes who were similar in how they accumulated MRI abnormality. Based on the earliest abnormalities suggested by our model they were called: cortex-led, normal appearing white matter-led, and lesion-led subtypes. In the external validation datasets, the lesion-led subtype showed a faster disability progression and higher disease activity than the cortex-led subtype. In all datasets, MRI-driven subtypes were associated with disability progression (βSubtype=0.04, p=0.02; βStage=-0.06, p<0.001), whilst clinical phenotypes and baseline disability were not. Only the lesion-led subtype showed a significant treatment response in three progressive multiple sclerosis randomised controlled trials (-66%, p=0.009) and in three relapsing remitting multiple sclerosis trials (-89%, p=0.04). Our results show that MRI-driven subtyping using machine learning can prospectively enrich clinical trials with patients who are most likely to respond to treatments.

Staging and stratifying cognitive dysfunction in multiple sclerosis

“…Even so, newer staging approaches do not rely upon the assumption of monotonicity and can also parse separate courses of progression, with potential for future application to cognition. 37,38 Future work may also wish to focus on specific biomarkers that precede or coincide with the cognitive impairment events outlined here. It would be especially relevant to explore how our cognitive model fits with previous models of grey matter atrophy progression in MS. 13 Additional biomarkers yet to be sequenced with cognition in this way, such as T2 lesion load or serum neurofilament light chain, may also shed light on the relationship between disease progression and cognitive decline.…”

Characterisation of MS phenotypes across the age span using a novel data set integrating 34 clinical trials (NO.MS cohort): Age is a key contributor to presentation

“…Advanced analytics and machine learning approaches have provided relevant new insights in-and outside of medicine and their utility needs to be explored in large MS data sets. There have only been a few attempts, [7][8][9][10][11][12] not surprising given the practical complexities of data access and the curating required. prognostic factors, using advanced analytical approaches applied on the novel NO.MS data set, composed of 34 Novartis MS clinical trials.…”

Machine and deep learning in MS research are just powerful statistics – Yes

“…In multiple sclerosis (MS) research, Machine and Deep Learning methods have been increasingly used, mainly to predict future clinical outcomes 4,5 or classify patients into prognostic groups based on patterns learnt from data. 6 Furthermore, these methods seem also very helpful for differential diagnosis. [7][8][9] Interestingly, most of these studies have applied Machine/Deep Learning techniques to neuroimaging or complex laboratory data, bringing to light the intrinsic complexity of these types of data, whose analysis centred only on conventional average-based metrics in the context of (classical) Statistics is possibly not powerful or compelling enough.…”