Programme for High-Dimensional Translation in Neurology (360G-Wellcome-213038_Z_18_Z)

Current models of the brain—in health and disease—are too simple to characterise individuals and to elucidate causal mechanisms of real-world complexity. Novel machine learning methods combined with large-scale data can render this complexity intelligible, enabling us to generate formalised, probabilistic, high-dimensional models with great individuating power. Enhanced individuation—personalisation—of patient care will not only improve clinical outcomes but also enhance healthcare productivity and illuminate pathological and interventional mechanisms otherwise opacified by unmodelled individual variation. The foundational framework we have established to enable high-dimensional modelling will deliver a suite of products addressing major unmet needs in stroke, neuroradiology, and delirium. Innovative features include: the development of compact, high-dimensional, deep-learning-based, widely-generalizable models of imaging and non-imaging data, the transformation of routine, heterogeneous, large-scale clinical data into analysable form, the unification of clinical, operational, and scientific objectives as a means of achieving early patient impact and accelerating adoption, and the deep integration of development pathways with healthcare delivery and enterprise. The broad platform our objectives will establish and validate will serve as the foundation for embedding high-dimensional modelling within healthcare, with future impact across the whole of medicine.