INNOVATIVE TOOLS FOR

DEMENTIA DIAGNOSIS

http://gtr.rcuk.ac.uk/projects?ref=EPM006093/1
 Impact Objectives
• Generate multivariate cognitive profiles in different dementias

• Develop novel ‘instruction-less’ tests of cognition suitable for patients

with different types and severities of dementia

• Enhance existing cognitive assessment paradigms using eye tracking

Innovative tools for Machine learning

dementia diagnosis applied to dementia
Professor Sebastian Crutch, Dr Neil Oxtoby and Emilie Brotherhood are part of a multidisciplinary
team committed to applying the same computational approaches currently used on neuroimaging
The C-PLACID project represents a major push by different centres at University College
data, to the assessment of cognitive impairment
London to apply machine learning and new cognitive assessment tools to the diagnosis of
physiology, cognitive NO: We have been able to blend and combine dementia to improve patients’ quality of life and inspire a new generation of researchers
assessment is long a number of different techniques and
overdue a step disciplines because we are in a rare position The incidence of dementia increases in Crutch of the Dementia Research Centre at more specific prognosis and care plan.
change in approach at UCL, with the co-location of researchers parallel with greater life expectancy. The UCL, believes that: ‘Cognitive assessment The C-PLACID project is bringing the same
akin to the many from many different fields. The project World Health Organisation predicts that is long overdue a step change in conduct, sophisticated analytical techniques into
advances in brain team sits at the interface between medical the number of people living with dementia precision and efficiency,’ and adds: ‘Many of the realm of cognitive assessment that
imaging we have engineering and the clinic. Our success is worldwide will triple by 2050 and the today’s tests fail to capture critical, sensitive are already applied to the interpretation of
Professor Sebastian Crutch Dr Neil Oxtoby Emilie Brotherhood
seen in recent years. based on the willingness of staff on both sides Alzheimer’s Disease Neuroimaging Initiative aspects of task performance and can be long neuroimaging data. Crutch explains: ‘As well
The C-PLACID of that professional and experiential divide estimates that it is the third most common and tiring for the patient to undergo.’ The as developing a computational platform for
Why is it important to improve tests for project brings together clinicians and to immerse themselves in another world. For cause of death for elderly people in the US. project aims to create and validate event- the analysis and visualisation of complex
cognitive impairment when assessing computer scientists to lay the foundations of instance, our computer scientists are meeting Being able to diagnose different forms of based models for dementia, improve current cognitive datasets, we are automating,
patients who have or may be at risk of such a change. people with different dementias and trying to dementia and to plot the evolution of disease tests and develop innovative cognitive optimising and creating techniques and
developing dementia? understand their experiences, for example, is an important component of tackling this assessment methods. devices for the acquisition of traditional
What types of testing do you advocate and at our regular support group meetings. problem. Early and accurate diagnosis, which and new cognitive data.’ The project will
SC: Cognitive impairment relates to changes in what way are these better than current Likewise, clinicians are embracing some of the leads to better management and improved TESTING FOR COGNITIVE IMPAIRMENT provide solutions to the many problems
in memory, perception, language, decision- methods? principles and advantages offered by machine- quality of life for patients, is the focus of The term ‘dementia’ covers a range of of assessment that have been highlighted
making and other mental skills associated learning. the three-year C-PLACID project that is conditions, including Alzheimer’s disease, by clinicians and develop tests that do not
with different forms of degenerative brain SC: We aim to combine the principles of developing a sophisticated computational dementia with Lewy bodies, the visual require instructions. The results are intended
disease. Dementia is generally defined traditional tests with new technology and What objectives has the project met so far? platform for the assessment of cognition. syndrome posterior cortical atrophy, to benefit both clinicians and patients,
as the point at which these changes start analytical methods. One example is our frontotemporal dementia, and vascular with the latter benefiting from shorter, less
to interfere with everyday life. There are eye-tracking analysis of spatial anticipation EB: Our first goal was to develop multivariate The project has brought together researchers dementia. Patients are referred to a specialist stressful testing and better management of
many different types of dementia and in patients with frontotemporal dementia. computational models for all types and from many disciplines within University once they notice signs of diminished memory, their specific condition.
understanding the cognitive profile of each Rather than merely focusing on the motor stages of different types of dementia. In this, College London (UCL), including computer language, reasoning or other cognitive
is critical to diagnosis, patient monitoring function of the eye, we are using eye motion the results have exceeded expectations and scientists, mathematicians, engineers, skills. In addition to potentially undergoing EYE TRACKING AND VIRTUAL REALITY
and evaluation of treatments. Many tests fail as an indicator of anticipative cognitive proven the potential of this type of machine- psychologists and dementia clinicians. neuroimaging, people under assessment are One of the innovative testing techniques
to capture critical, sensitive aspects of task ability. Our initial results indicate that the learning to predict the progression of disease, Principal investigator, Professor Sebastian commonly subjected to a battery of cognitive being developed by the team is the
performance such as naming tests which fine-grained, rich datasets obtained from eye- based on different events which signify tests. However, Crutch explains that these design of instructionless eye-tracking
measure the accuracy with which someone tracking measurements can inform us about notable changes in condition. The second set tests have several limitations: ‘Many of today’s tests of cognition. In one experiment,
can retrieve a word but miss other sensitive high level cognitive functions in dementia. of goals is to improve cognitive tests using tests do not have the sensitivity to distinguish patients with different forms of dementia
performance indications such as how long it By using machine-learning we can identify novel technologies such as automatically between different types of cognitive are given a computerised version of the
takes to do so. The tools we currently have to conditions where subtle deficits are present measuring voice reaction times and tracking impairment and results can be affected by Brixton spatial anticipation test. This test
assess cognition are important and valuable, and thereby contribute to test optimisation eye movement in a spatial anticipation task. the ways in which each clinic carries out shows patients a matrix of white circles
but many have changed little in the past and the reduction of testing times. Such tests Over 850 people have undergone eye tracking the assessment. Some require complicated through which a black dot traces patterns
three or four decades and are often long to also avoid the need for instructing patients on and other tests are being implemented on instructions and there are domains of that are followed by the eye, with the
administer and tiring for people undergoing how to complete the tasks, which rules out 500 of the 1946 British-birth cohort during its cognition that they cannot assess.’ pattern changing unpredictably. Patients
assessment. With the advent of a slew of mistakes caused by misunderstandings or second phase of neuropsychological testing. are normally required to give a verbal
computational approaches for analysing language difficulties. Other tasks are underway, including the use It is vital to identify the specific condition indication as to where the dot will appear,
complex datasets, greater computational of virtual reality to assess ability to function in and the stage of progression of a patient’s whereas in the computerised version, the
A scene from the virtual cafe bistro in which immersed
power and new tools for evaluating What has the key to your project’s success to different social settings. disease. Pinpointing these markers with eye motion is automatically tracked. Trials
participants observe and interact with avatars in
different aspects of human psychology and date been? different social scenarios accuracy enables the development of a were conducted on 12 behavioural variant
 Cognitive assessment is long overdue a step
change in conduct, precision and efficiency

frontotemporal dementia patients (bvFTD) movements is yet to be perfected. Ideally, advances that will facilitate the gathering of
who were expected to show deficits in this both would be delivered via a head-mounted more cognitive data, such as body-mounted
test; six semantic dementia patients who device (HMD), but Brotherhood notes: ‘Eye sensors and improved ways of synchronising
were expected to show less impairment tracking needs to be available in a HMD and different inputs.
in undertaking this task and 38 healthy it would certainly make testing much simpler.
controls. The eye movement measurements In addition, we could incorporate sensors
were studied using both frequentist to measure head movements towards or
statistical analysis and machine-learning away from a stimulus, which would add even
algorithms trained on healthy datasets. greater depth to the test data we can gather.’
The machine-learning approach performed
far better than both the original pen and INSPIRING MORE DEMENTIA RESEARCH Project Insights
paper testing of the same patients and The C-PLACID team is hopeful that their
the results from the frequentist statistical work will improve diagnosis and treatment of FUNDING
This work was supported by the
analysis. Crutch says: ‘It reported significant individuals with dementia and inspire other
Engineering and Physical Sciences
differences from controls in all the trials scientists to work in this field. As Crutch says:
Research Council (EPSRC) EP/M006093/1.
conducted on bvFTD patients and in half ‘By elevating the assessment of cognition to
of the SD patients, whereas both other the same level as neuroimaging, we imagine CONTACT
approaches missed deficits in a significant researchers may begin to develop new Professor Sebastian Crutch
proportion of participants.’ measures of cognition and to study how all Project Coordinator
these aspects of cognitive performance relate
T: +44 (0)203 448 3113
A number of other novel testing techniques to one another.’ Crutch cites an example from
E: [email protected]
are being developed and tested by the team. their own work: ‘We have noted that when
W: http://gtr.rcuk.ac.uk/projects?ref=EP
Mark Huckvale, a speech scientist from UCL, testing overt behavioural responses, we have M006093/1
has helped create a means to complement also been able to record some covert aspects
picture-naming tests with measures of of performance, such as cognitive effort. This BIO
response speed. 500 members of the 1946 can give us information on how a patient Sebastian Crutch is a neuropsychologist
British-birth cohort are currently undergoing is reacting to the experimental test setting and professorial research associate at the
Dementia Research Centre, UCL Institute
this test to determine if naming speed is and enable adaptations and standards to be
of Neurology. His research focuses on rare
indicative of the build-up of proteins in the developed.’
and young-onset dementias, exploring
brain associated with Alzheimer’s disease. topics such as dementia-related visual
There are ethical and privacy considerations Dr Neil Oxtoby, a research scientist from impairment, computational approaches
to be considered, but the trial showed that the Centre for Medical Image Computing to improving cognitive assessment, and
important cognitive data can be mined from at UCL, sees the C-PLACID computational reading and balance problems. Currently,
the way the patient engages with the device. approach being applied more widely: ‘We he directs the Created Out of Mind 2016-
2018 dementia and arts residency at The
believe that computational platforms
Hub, Wellcome Collection, which aims to
Emilie Brotherhood, a research scientist at for quantitative assessment of disease
shape and enrich public and professional
the UCL Dementia Research Centre, has progression are the next step for modern perceptions of the dementias and explore
been involved in the use of virtual reality healthcare across a whole range of diseases.’ the opportunities afforded by collaborative,
to expose patients to a scenario in which Large amounts of digital data are now interdisciplinary, publicly-situated research.
social cognition can be assessed. She says: available for many conditions and thus can
‘People with frontotemporal dementia, be interrogated using machine learning and
for instance, may experience a myriad of artificial intelligence platforms. He notes:
changes in the way they perceive, behave ‘All the big IT companies are involved in
and interact in different social settings. developing systems for analysing digital
Finding a social scenario that is relevant health records and providing data-driven
to all patients has been a real challenge.’ patient care in hospitals. C-PLACID is part of
The technology available for delivering a that movement.’ Brotherhood is also hopeful
virtual reality experience and for tracking eye that their work will lead to technological