# Predictive Prognostic Models for predicting cognitive change in the context of multiple long-term disorders This project will develop and end-to-end pipeline for automated prediction of cognitive decline using low-cost non-invasive clinical data in people with multiple long-term conditions attending memory clinics. We take advantage of standard clinical assessments of a structural MRI scan and a standard brief cognitive assessment (Revised Addenbrookes Cognitive Examination). These data will be used for individualised trajectory modelling of cognition which can be used to apply clinical pathways, and potentially for identifying individuals suitable for clinical trials. For modelling the trajectory of cognition we will employ a Predictive Prognostic Model (PPM) that we have developed (Giorgio et al., 2020). This model has been successful in predicting cognitive decline in Mild Cognitive Impairment and Alzheimer’s disease in the Alzheimer Disease Neuroimaging Initiative (ADNI). Significantly, the model has been validated in independent international cohorts (eg NIMROD, UK; BACS, USA). Further development since publication of the model has used pilot data from the Addenbrookes Memory Clinic in a proof of concept to translate the PPM into a clinical workflow. The prototype pipeline is based in XNAT for clinical imaging data and calls cognitive and diagnostic data from a REDCap database. The pipeline: * Automates the pre-processing of MRI scans * Integrates clinical brain scans with cognitive tests * Extracts key markers that are predictive of disease * Implements PPM to derive disease progression trajectories * Returns individualised prognostic scores via a browser-based user-interface to support clinical management