SAFEVID-TREs: Spasms Analysis using FEderated learning from VIDeos across multiple TREs

Neurological conditions affect approximately 11 million people in the UK, creating a major public health crisis. They are a leading cause of disability, responsible for about 140,000 deaths annually (one-fifth of all UK deaths), and account for 800,000 hospital admissions each year.
A critical issue is the significant delay in diagnosis and intervention. For children and young people (CYP), this delay often exceeds 12 months for 30% of cases. Specifically:

28% wait over 7 months just to see a paediatrician after their first GP visit.
36% wait over 7 months to consult a neurologist after seeing a paediatrician.

This delay leads to poor symptom control, repeated unscheduled hospital visits, and in-patient care. Conservatively, the cost of neurological disorders to the UK economy is estimated at over £100 billion per year. Early identification and intervention are challenging but are vital for improving personal health outcomes and easing pressure on healthcare services.
Our interdisciplinary team is actively addressing this challenge by developing machine learning models that analyze infant movement videos. Our focus is on predicting specific neurological conditions, such as:

Detecting Infantile Epileptic Spasms (IES)
Automating the General Movement Assessment (GMA) for the early prediction of Cerebral Palsy

We have already secured over £560K in research funding to develop AI prototypes within a secure Trusted Research Environment (TRE).
While the necessary environment for training AI models in our current TRE is established, the lack of federated learning capabilities has created a critical bottleneck. This limitation restricts our access to diverse data from different patient cohorts, which hinders our ability to enhance the generality and robustness of the AI model.
The new TREvolution programme introduces simplified federated learning capabilities within TREs, addressing both the technical and governance hurdles. Leveraging this innovation, our...