Globally, 50 million people live with dementia and health and social care costs are estimated to be £26 billion in the UK alone. Individualised diagnosis at early stages of decline in cognitive abilities has the potential to increase the success of interventions and treatment. This will reduce pressure on global care systems. Yet, patients often have other conditions at the same time, which makes diagnosis difficult. Here we consider the case of late-life depression as key comorbidity (that is, occurring at the same time) of memory decline and dementia diagnosis. We aim to develop machine learning approaches that use several types of health data, including structural and functional brain scans and results from cognitive tests, to discriminate between dementia and depression. We will use data collected over a long period of time to predict whether individuals are at risk for dementia and/or depression based on genetic risk. This approach will reduce the risk of patient misclassification and provide tools for organising patients into subgroups for which tailored and effective treatments can be developed. Our approach has the potential to inform patient selection for clinical trials. Furthermore, it can reduce healthcare costs by assigning the right patients to the right diagnostic and treatment pathways.
