AI for better brain and mental health

<h3>Biography</h3> Zoe Kourtzi is Professor Computational Cognitive Neuroscience at the Department of Psychology, University of Cambridge. Her experimental work aims to understand the role of lifelong learning and brain plasticity in enabling humans of all ages to translate sensory experience into adaptive behaviours. Her computational work aims to develop AI-guided predictive models of brain and mental health. Her work has translational impact in the early diagnosis and design of personalised interventions in brain and mental health. Zoe received her PhD from Rutgers University and was postdoctoral fellow at MIT and Harvard University. She was a Senior Research Scientist at the Max Planck Institute for Biological Cybernetics and then a Chair in Brain Imaging at the University of Birmingham. She moved to the University of Cambridge in 2013 and she is the Angharad Dodds John Fellow at Downing College. She is a Royal Society Industry Fellow, Fellow and Cambridge University Lead at the Alan Turing Institute and Co-director of Cambridge’s Centre for Data Driven Discovery. Early prediction of neurodegenerative disorders is key for clinical management and patient outcomes. Alzheimer’s disease (AD) is the commonest type of dementia, and is characterised by progression from normal cognition, to mild cognitive impairment (MCI), to dementia. However, not all individuals with MCI develop dementia. Predicting whether individuals with (MCI) or older people without symptoms will decline or remain stable is impeded by patient heterogeneity due to factors such as comorbidities, lifestyle and disease severity. Despite the importance of early diagnosis of AD for prognosis and personalised interventions, we still lack robust tools for predicting individual progression to dementia. We propose a novel AI-guided predictive prognostic modelling (PPM) approach that mines multimodal data to derive an individualised prognostic marker of cognitive decline at early stages of dementia or before symptoms occur. We validate our approach against routinely collected real-world patient data from memory clinics, enhancing clinical utility and translation to clinical settings. Our approach has strong potential to facilitate effective stratification of individuals based on prognostic disease trajectories, reducing patient misdiagnosis with important implications for clinical practice and discovery of personalised interventions.