Research Questions
Can we create an artificial intelligence (AI) model that faithfully captures complex individual variations in biological characteristics related to the nervous system, and distinguishes normal and measurement-related variations from those associated with disease?
Can the model improve our ability to diagnose common neurological diseases, predict their progression, and recommend treatments?
Research Objectives
Our primary objectives are:
To develop a type of AI model-a “foundation” model-that learns the relationships between multiple types of brain-related data in the UK Biobank dataset.
To test the model’s utility for diagnosing nervous system disorders, predicting their progression, and identifying responses to treatment in individual patients.
Scientific Rationale
Disorders of the nervous system are extraordinarily complex. When we examine brain scans and other medical tests, it is often difficult to distinguish characteristics caused by disease from those that occur naturally or arise from imperfect instruments. Current approaches examine each type of test separately, missing important connections between different measurements. Our solution is to develop a novel AI-based “foundation” model that learns from all types of nervous system-related data together, enabling the recognition of test patterns that distinguish disease-related changes from normal variation. Recent advances in algorithms and hardware have made such models feasible. The proposed model could support more accurate diagnosis by integrating data across many tests; fairer care by accounting for population-level variation; and new insights into disease mechanisms, supporting personalised treatments tailored to individual characteristics. This represents a shift from looking at isolated data to understanding the brain as a system, potentially transforming diagnosis and treatment of neurological conditions.