Skip to main contentSkip to footer
Telescopic image of a large orange and blue galaxy
Brighton & Sussex Medical School



The ASTRODEM project aims to create a predictive model which will help general practitioners (GPs) identify patients at high risk of dementia.

University of Sussex astrophysicists will swap galaxies for general practice patient data in an innovative new study, in collaboration with researchers from Brighton and Sussex Medical School.

ASTRODEM is funded by a Wellcome Trust Seed Award in Science.

Find out more about the project by watching the video, made by the Wellcome Trust, below.


Telescopic image of a very large purple spiral galaxy


We aim to create a predictive model which will help general practitioners (GPs) identify patients at high risk of dementia. We are using 96,000 anonymised GP patient records from the Clinical Practice Research Datalink, to train our model. Using modelling techniques drawn from physics and astronomy, our project is different from others in the following ways: 

  1. We are using imaginative methods, drawn from a range of disciplines, for identifying and selecting of predictors
  2. We are using information from the longitudinal nature of the patient pathway, represented in the patient record
  3. We are accounting for high rates of misdiagnosis and other noise in the data
  4. Making no assumptions about the data, our team is also surveying GPs to understand why GPs do and don't pursue a diagnosis of dementia with different patients, and how GPs record uncertainty about presentations of the condition.

Watch a video of Dr Elizabeth Ford presenting 'ASTRODEM: Using Astrophysics to Close the Diagnosis Gap for Dementia' at Future Medicine 2017 in Berlin 

A set of yellow and black brain scan images

Why dementia?

Dementia is one of the greatest public health challenges of our era. There are an estimated 800,000 people with dementia in the UK and by 2021 this will increase to over one million. The impact of this disorder on patients, their carers, other family members and society is profound. Timely diagnosis allows patients to benefit from current therapies, plan for the future, and maximise their quality of life. However, there is a 'diagnosis gap' in UK general practice, with less than two-thirds of expected patients receiving a dementia diagnosis. Currently diagnosis often occurs late or not at all and opportunities for therapeutic intervention are missed.  

Higher diagnosis rates in general practice, and diagnosis earlier in the disease course, are strategic aims for the UK government and NHS, set out in the National Dementia Strategy, Prime Minister’s Dementia Challenge, NHS Innovation Challenge Prize for Dementia, NHS England Dementia Identification Scheme and the GP Dementia Toolkit.

Telescopic image of a purple and white spiral shaped nebulea

Why astronomy?

Astrophysicists have long faced many challenges that are only newly emerging in health big data, as they interrogate catalogues of galaxies and galaxy spectra. Astrophysicists at the University of Sussex have established a library of statistical techniques for specifying and identifying galaxy 'finger-prints' and these same methods can enable epidemiological and health services researchers to exploit routine health data more effectively. Physicists seek optimal techniques by applying methods from a diverse range of fields, allowing them to identify the same galaxy seen in pictures from different telescopes (data linkage), to find sets of galaxies that have similar properties and identify the common features of galaxy 'finger-prints' (case-definition) and to create models of galaxies that predict some behaviour on the basis of other observations (predictive diagnostics). These techniques can be used on GP patient records, where the clinical condition is analogous to a galaxy, the GP database to a galaxy catalogue, and diagnosis of a condition to the modelling of galaxy properties. 

Astrophysicists work with blurred or noisy data and can bring a range of novel techniques to address data quality challenges, missing data, and uncertain classifications within the training data set.