Machine learning to optimise DECODE dementia identification clinical software


Principal investigator(s)Co-InvestigatorPhD student or Research Fellow(s)Project title
Dr David Llewellyn

Professor William Hamilton, Professor Richard Everson, Professor Edward Keedwell, Dr Sarah Moore, Professor Jonathan Fieldsend

Dr Diane Fraser, Dr Ben Evans Machine learning to optimise DECODE dementia identification clinical software


Lay summary:

Many people with dementia are never diagnosed or are diagnosed during the later stages of the condition when a diagnosis may be less helpful. A substantial investment was therefore made in setting up a network of specialist memory clinics across the UK to which patients could be referred by their general practitioners (GPs). However it is a difficult task for GPs to recognise dementia symptoms and respond appropriately in the limited time available. Currently only around half of patients referred to memory clinics have dementia. Waiting times are rapidly increasing as there is simply not enough capacity to see everyone.

With the support of the Halpin Trust charity we have developed a DEmentia identification COmputerized DEcision support system (DECODE). We know that our system is considerably more accurate than the current ‘gold standard’ assessment. However new advances in artificial intelligence mean that it should be possible to improve the accuracy of the system even further. Indeed recent publications demonstrate that new approaches to analysing data have the potential to be even more accurate than doctors in detecting conditions such as cancer. The focus of this project is to provide the results that we need to secure major funding to undertake a bigger project. Our ultimate vision is to develop an intelligent system that helps doctors identify people with dementia and reduces unnecessary assessments, benefitting both patients and the NHS.