Computational modelling of care needs in the rare dementias

Very little information is available about the order of appearance and rates of change for symptoms and care-related needs in the rarer dementias. When the "What next?" question is raised, people living with dementia (PLWD) and carers all too often receive responses such as "It is difficult to say" or "It affects everyone differently" - which are simultaneously true and completely unhelpful.

Approximate guides to the stages of Alzheimer's disease have been produced (e.g. Reisberg's 7 stages of Alzheimer's) but these are based on clinical experience, not quantitative data, and are not available for most rare, atypical or young-onset conditions.

The overarching goal of the project is to provide doctors, patients and families with tools to better plan for people's future - finally answering the "What next?" question.

Data-driven quantitative approaches to understanding neurological disease progression emerged early in the last decade, benefitting from the availability of large medical data sets in neurodegenerative diseases such as Alzheimer's disease, the most common cause of dementia. "Imaging Plus X" computational models are now a mature research field in their own right [Oxtoby2017; Khatami2019]. One of the original methods, the event-based model (EBM [Fonteijn2012; Young2014]), is a robust computational approach to describing disease progression without requiring a priori clinical staging and requiring only cross-sectional data.

EBMs have been applied across a range of neurological diseases including Alzheimer's disease, Huntington's disease, and Multiple Sclerosis, using biomarker-based events including regional atrophy (shrinkage) in the brain, levels of abnormal proteins in cerebrospinal fluid, and more recently in work on measures of cognitive ability [Firth2019; Firth2020]. While this work has generated unique biomarker-based understanding of disease progression, it tells us very little about patient and carer experience.

This project will address this gap by developing and applying data-driven computational modelling and machine learning techniques to self-report data from PLWD and their carers.