MS is divided into four subtypes: clinically isolated syndrome, relapsing remitting, primary and secondary progressive MS and are used to guide the timing and choice of treatments. However, these subtypes are based on observed symptoms, such as relapses and disability which can be difficult to measure and may not reflect the underlying biology driving the course of someone’s MS.
Researchers wanted to find out if there are hidden patterns in MRI brain scans taken over time that would better identify biological differences in MS activity and detect progression earlier.
The international team used MRI scans previously taken in clinical trials involving 6322 people with MS. Data was extracted from the scans and machine learning (artificial intelligence) used to identify subgroups with similar patterns of change in brain structures over time. Results from the initial findings were tested against a second set of MRI scans from 3,068 people with MS.
The researchers found they could classify people into three subtypes, based on early changes seen on MRI scans:
- Cortex-led subtype showed early signs of tissue shrinkage in the outer layer of the brain
- Normal-appearing white matter-led subtype began with diffuse tissue changes in the middle of the brain
- Lesion-led subtype started with widespread build-up of lesions, followed by early and severe shrinkage of brain tissues in several areas.
The researchers conclude that the three new MRI-based subtypes are better able to predict MS activity, disability progression and treatment response than the standard relapsing and progressive subtypes.
The findings have the potential to open up research into new, more effective, treatments by enabling researchers to select clinical trial participants based on the biology which is driving the course of their MS. In the future, this approach could be used in the clinic to support treatment choices.
Historically, MS has been divided into four subtypes: clinically isolated syndrome, relapsing remitting, primary progressive and secondary progressive MS. These have been used to guide the timing and choice of treatments for MS. However, these subtypes are based on observed symptoms, such as relapses and progression of disability which can be difficult to measure accurately and may not reflect the underlying biology driving the course of MS.
Researchers wanted to find out if there are hidden patterns in MRI brain scans taken over time that would better identify biological differences in MS activity and detect progression earlier. This approach might give a better guide to treatment choice and identify people who would respond to a particular treatment.
The international team which included researchers from University College London, collected MRI scans previously obtained from 16 clinical studies involving 6322 people with MS (training set). These included scans taken from four primary progressive, five secondary progressive and three relapsing remitting studies. Data was extracted from the scans and machine learning (artificial intelligence) used to identify subgroups which had similar patterns of change in brain structures over time.
Results from the initial findings were tested against a second set of MRI scans (validation set) from 3,068 people with MS. These included scans from one primary progressive, two secondary progressive and two relapsing remitting studies.
The researchers found they could classify people into three subtypes, based on early changes seen on MRI scans:
- Cortex-led subtype showed early signs of tissue shrinkage in the outer layer of the brain
- Normal-appearing white matter-led subtype began with diffuse tissue changes in the middle of the brain
- Lesion-led subtype started with widespread build-up of lesions, followed by early and severe shrinkage of brain tissues in several areas.
The cortex-led subtype was most common in both sets of MRIs, with the normal-appearing white matter-led second most common in the training set, and the lesion-led subtype second most common in the validation set.
The lesion-led subtype had the highest relapse rates and risk of disability progression but was also the subtype most likely to show a response to treatments for both relapsing and progressive participants in clinical trials.
Combining MRI-based data with clinical assessments taken at the start of clinical trials (EDSS, timed walking test or 9-hole peg test for upper limb function) gave a more reliable prediction of future disability progression.
The researchers conclude that the three new MRI-based subtypes are better able to predict MS activity, disability progression and treatment response than the standard relapsing and progressive subtypes.
This study was based on a review of MRIs previously taken for clinical trials. The next step will be to confirm the findings in clinical studies which select participants from the outset based on their MRI subtypes. The researchers also noted that the data used for this study did not include scans of the spinal cord as these are not routinely collected in clinical trials; future studies could investigate the effect of adding spinal cord measures to MRI subtypes.
The findings have the potential to open up research into new, more effective, treatments by enabling researchers to select clinical trial participants based on the biology which is driving the course of their MS.
Further research will be needed to translate the findings into practical guidelines that can be used in the clinic to help predict those who are more likely to have disease progression and support treatment choices for those who would best respond to a particular therapy.
Eshaghi A, et al.
Identifying multiple sclerosis subtypes using unsupervised machine learning and MRI data
Nature Communications 2021; 12(1): 2078.
Full paper
Magnetic resonance imaging (MRI) is a scanning technique used on the brain and/or spinal cord support MS diagnosis and to monitor on-going MS activity. You can read more about the MRIs and what it’s like to have an MRI scan.
From our Open Door archive, this blog offers a different insight into what MRI scans show.
