MS-PINPOINT · Research programme
Our work runs from measuring the brain and spinal cord on ordinary clinical MRI, through predicting and explaining how MS progresses, to designing the trials and the privacy-preserving infrastructure that let it all happen at scale, across hospitals internationally.
The programme
Filter by theme or search to find a strand, then open it to read what it involves and where it connects.
8 research areas
Prediction & causality
Prognosis — predicting the future course of the disease — is where imaging, clinical history and machine learning meet. We combine markers from brain and spinal-cord MRI with routine clinical data to estimate how an individual's MS is likely to progress, so that monitoring and treatment can be matched to risk rather than to the average patient.
Prediction & causality
Most machine learning finds patterns; it does not tell you what would happen if you intervened. Causal machine learning does. We use it to distinguish the factors that genuinely drive MS progression from those that only correlate with it — the foundation for estimating how a given treatment would change a given person's course, rather than the population average.
Related
Imaging & biomarkers
Research-grade image analysis usually assumes research-grade scans. Real hospital archives are heterogeneous — different scanners, contrasts and resolutions. We build machine-learning methods that measure lesions and brain and spinal-cord atrophy robustly across that variety, turning routine clinical MRI into quantitative biomarkers of disease activity and progression.
Imaging & biomarkers
The current subtypes of MS were defined by clinical observation decades ago and only loosely track the underlying biology. Using data-driven models over imaging and clinical data, we are working towards a classification grounded in the mechanisms that actually differ between people — one that could better predict progression and response to treatment.
Related
Clinical trials
Trials in progressive MS are long and expensive, partly because the outcome measures are noisy. We develop more sensitive, imaging-derived measures of progression and use them to improve how trials are designed and powered — so that effective treatments can be identified with fewer patients and less time.
Related
Clinical trials
Conventional trials systematically under-include people who are unable to travel — often those with the most advanced disability, exactly the group progressive-MS research most needs to reach. We work on remote and decentralised trial methods that let people take part from home, widening who gets to be studied and who benefits.
Related
Data & network
Working with hospital imaging at scale only earns trust if privacy is built in. We use privacy-preserving approaches such as federated learning — where models are trained across many hospitals without patient data ever leaving the site that holds it — alongside strong de-identification, so research can span many centres while records stay local.
Data & network
None of this works from a single site. We build and sustain a network of hospitals and academic partners internationally, and — through our advisory board and patient involvement — keep people affected by MS at the centre of what we choose to study and how.
Related