Prediction Models of Dementia Risk from Brain MR Images and Proxies of Cognitive Reserve

Supervisors: Professor Alison Murray (University of Aberdeen), Dr Roger Staff (NHS), Professor Joanna Wardlaw, Professor Craig Ritchie (University of Edinburgh) and Dr Robin Wolz (IXICO)

The Aberdeen Children of the Nineteen Fifties (ACONF) cohort provides rich perinatal and early life data and, at age 59–65, is ideal for longitudinal dementia research due to the participants’ proximity to the risk period for cognitive decline and dementia. A sub-sample are currently being recruited as part of a Wellcome Trust funded study for clinical, cognitive and brain MRI assessment, to which we will add. Using structural and functional brain imaging in targeted sub-samples we will examine life-course determinants of accumulation of disease burden associated with cognitive impairment and dementia. While much can be measure from brain MRI, we cannot currently diagnose cognitive impairment or dementia from a brain scan which may be due to individual variability in cognitive reserve or resilience. This project will develop models of individual resilience and determine which key life variables allow us to predict dementia risk from brain images, providing a basis for future research and clinical trials. We will initially use standard brain MRI variables that are visually and quantitatively extracted from structural images (atrophy, WMH score, microbleed score, voxel-based volumes to predict cognitive outcomes. We will then employ a machine learning approach, in collaboration with industry partners IXICO, using key life-course, cognitive reserve proxies and imaging variables to predict cognitive decline and dementia risk. That is, given the life-course of an individual, what are the imaging characteristics that allow us to predict cognitive decline and therefore future dementia risk?