Precision Treatment Strategies in Multiple Sclerosis with AI
Research type
Research Study
Full title
Precision Treatment Strategies in Multiple Sclerosis Using Next-generation Machine Learning
IRAS ID
319964
Contact name
Arman Eshaghi
Contact email
Sponsor organisation
UCL
Duration of Study in the UK
5 years, 2 months, 1 days
Research summary
Background
In multiple sclerosis (MS), the immune system attacks the brain and spinal cord. This causes problems in how we
move, feel or think. In the UK, 130,000 people live with MS, costing the NHS more than £2.9 billion a year.
MS types are based on when and how symptoms appear. Most cases (80%) begin as relapsing-remitting MS, with
unpredictable attacks and remissions. After 10-20 years, about 50% of patients develop secondary progressive
MS with gradually worsening symptoms without obvious attacks. In primary progressive MS, symptoms slowly
worsen from the start and accumulate over many years.
This classification does not reflect the underlying tissue changes (pathology), potential response to treatment or
future disability. More than 15 treatments currently exist, each with different risks and strengths. Brain images,
such as magnetic resonance imaging (MRI) and blood tests (serum neurofilament levels [sNfL]), help identify the
underlying tissue changes. If types of MS were redefined based on the underlying tissue changes, doctors could
recommend the most appropriate treatment with the fewest side effects.
Artificial intelligence (AI) uses computers for tasks that otherwise rely on human intelligence. I have developed an
AI tool to detect MS types based on MRI from clinical trials. However, clinical trials have strict criteria for taking
part. They typically exclude people in certain age ranges or who have more than one lifelong condition. I now want
to develop new tools for real-world use from diverse patient populations so that my tools can benefit anyone with
MS.
Aim
I will identify patterns of MRI and serum neurofilament (where available) change that characterise specific types of
MS. These types will predict how symptoms may change, when MS may worsen, and which drugs are most likely
to help.
Methods
I will work with colleagues from seven NHS Trusts and two international clinical centres to:
1) Use MRI and, where available, sNfL to define MS types that reflect underlying abnormalities better than
symptom-based types.
2) Build AI tools that measure and predict whether a treatment works at an individual level. This will make it
easier to tell how new MS types predict treatment effects in NHS data.
3) Build AI tools based on objectives (1) and (2) that can learn from each other without sharing patients’ data
across centres to optimise and refine the predictions. This ensures that I can include more patients from diverse
backgrounds for truly personalised predictions.
I will measure how new MS types complement the existing MS types based on disability outcomes and calculate
how my digital tools may save costs in future healthcare by recommending appropriate treatments earlier and
delaying disability.
Benefits
The result will be research tools that predict whether a treatment works using widely available NHS data. It will
give researchers information on the probability of treatment response and disease worsening. In the future, this
could evolve into software that will be used in hospitals to help doctors recommend treatments, minimise disability
accrual, reduce side effects and save money.
Patient and public involvement (PPI)
This proposal was developed in consultation with a PPI steering committee. The committee will meet twice a year
to continue to shape the research and advise on the acceptability of the future medical device software tool,
interface, risks and benefits. The committee will assist in designing a follow-on path for future use in clinical
practice.
Dissemination
In addition to scientific articles and congress presentations, I will produce lay-language Q&A video summaries with
the PPI committee. These will be published on a dedicated website and MS patient forums identified by thesteering committee (e.g., MS Trust and Society) and promoted on social media.REC name
West of Scotland REC 5
REC reference
23/WS/0008
Date of REC Opinion
24 Jan 2023
REC opinion
Favourable Opinion