21 September 2022
Using CANTAB to develop AI-based tools for the early detection of dementia
To mark World Alzheimer's Day, we hear from Christoffer Hatlestad-Hall and his team at Oslo University Hospital, who explain how CANTABTM is an instrumental part of their study using AI and machine learning to identify the earliest signs of Alzheimer's and dementia in people with mild cognitive impairment.
Can you tell us about yourself?
My name is Christoffer Hatlestad-Hall, and I am a member of the research staff at the Department of Neurology, Oslo University Hospital, Norway, and co-founder of the medical technology start-up BrainSymph AS. Originally a clinical psychologist, my research interests are centred on the association between cognitive deficits in neurological patient populations and functional brain network characteristics. I am also a member of the AI-Mind project team, where I work with everything from neuroscientific methods and technology development to supervision and clinical study administration.
Can you tell us more about the research?
AI-Mind is a 5-year project funded by the EU under Horizon 2020 that kicked off on 1 March 2021. It brings together fifteen partners from eight countries, including academic institutions, medical centres, SMEs and patient organizations. With this project, we aim to develop two artificial intelligence (AI)-based tools for dementia prevention. The AI-Mind Connector will identify dysfunctional brain networks and the AI-Mind Predictor will assess dementia risk using data from the Connector, advanced cognitive tests (CANTABTM) and genetic biomarkers. These two tools will be integrated into an intelligent diagnostics platform to identify both brain network disturbances and dementia risk.
Why is the AI-Mind research important?
According to the World Health Organization (WHO), dementia affects approximately 55 million people around the globe. Despite the societal and economic burden associated with the disease, there is still a huge gap in our knowledge when it comes to identifying the early stages of its development. Having knowledge of the earliest changes in dementia will help with understanding how to prevent it.
More than 10 million Europeans show signs of mild cognitive impairment (MCI), a condition in which the person experiences some memory loss and problems with concentration and/or thinking but maintains the ability to independently perform most activities of daily living. It is estimated that around 50% of people with MCI will develop dementia within five years1 and the risk is more than 20 times higher than in the healthy elderly population. Current clinical practice (e.g. diagnosis, examination) lacks the necessary screening tools to identify which of those 50% are the ones at risk. The patient’s journey typically takes many years and involves several clinical visits before a conclusive diagnosis of dementia is finally reached. We will radically shorten this journey, through a digital AI-Mind solution that will provide a fast and accurate prediction for an individual’s dementia risk. This would give doctors and patients a window for preventive interventions, therapies, and rehabilitation measures early in the course of the disease.
Tell us a bit about the research and methods used
The AI-Mind study aims to enrol 1000 participants with MCI aged between 60 and 80 at five clinical centres in Norway, Italy, Finland and Spain. To get enrolled in the AI-Mind study, each person goes through the two step screening inclusion process: pre-(self-) screening and professional screening. People who report mild memory impairment symptoms through a screening questionnaire in the first step are then invited for the second step. In our professional screening procedure, our neuropsychologists and neurology specialists use standardised assessment tools that include a multi-domain neuropsychological test battery, Mini Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), Clinical Dementia Rating (CDR), and Instrumental Activities of Daily Living (IADL).
The study entails four visits over two years. At each visit, the participants undergo an electroencephalography (EEG) session and cognitive testing with the CANTABTM software on iPads. Next to that, at the first visit, participants undergo a detailed interview with a team member and a blood sample is taken for analysis.
The idea behind the AI-Mind initiative originates from the evidence that before structural brain pathology occurs in dementia, the brain exhibits subtle functional changes in its network activity. These changes can be reliably indexed through graph mathematical concepts applied to measures of functional connectivity in the EEG. This is highly important, as recent research demonstrates a significant link between such measures and cognitive functioning2,3. AI-Mind is the first large-scale study to address this relationship.
In addition to the EEG and cognitive data, in AI-Mind our researchers collect a blood sample for APOE allele genotyping and p-Tau analyses, and from two of our clinical sites (HUH and UCM), additional magnetoencephalography measures (MEG).
We will combine all these data sources using complementary machine learning and deep learning approaches to create a model that may predict, with high specificity and sensitivity, if a person with MCI is at risk for developing dementia. Based on the literature consensus, we expect that approximately half of our enrolled participants will progress from MCI to dementia during their two-year follow-up. The important question is whether the AI-Mind tools can predict who will develop dementia and whether we can capture this information at the earliest possible time.
PhD students Emily Beuken, Thomas Tveitstøl, and Mats Tveter work on the AI-MIND project
What are your key findings?
AI-Mind is still in its data collection phase and so far, the study has attracted more than 400 people with MCI. At the same time, the interdisciplinary AI-Mind consortium is working hard in a wide range of fields, from artificial intelligence (AI) and software development to health technology assessment and data protection regulations. Ongoing tasks include the active development of the AI-Mind platform, which will host the Connector and Predictor tools, user interface and experience design, and AI algorithm testing, among others. The talented researchers of the consortium, both the experienced and the fresh PhD students, are eagerly waiting to delve into the data at the first possible occasion.
Why did you choose CANTABTM for your study?
In large multi-site studies, such as AI-Mind, standardisation is crucial, and cognitive data in particular are prone to biases caused by inconsistencies in how cognitive tests are administered at different sites.. The innovative CANTABTM cognitive tests are very useful in mitigating these challenges. The tests are the same across countries and easily available on a tablet. Furthermore, the purely computerised administration, testing, and scoring of CANTABTM greatly ease the process of storing and curating cognitive data. With the expert assistance of the CANTABTM technical team, the AI-Mind data are being automatically pushed to our secure project server on a daily basis.
The CANTABTM battery employed in AI-Mind contains nine well-validated cognitive tests, all available and administered in the native languages of our partner sites (Norwegian, Spanish, Italian, Finnish and Swedish). Among the CANTABTM tests we use are the Pattern Recognition Memory (PRM) and Paired Associates Learning (PAL). The former measures memory in an efficient manner using randomized abstract stimuli, to minimise any learning effects and cultural bias, while the latter is known to be sensitive to cognitive deterioration. The documented language and culture-independency and high sensitivity/specificity makes CANTABTM suitable for use in an MCI population which may range considerably in terms of their cognitive challenges.
Conducting a clinical study that involves many partners and people requires flexibility in terms of administration. Fortunately, managing test batteries, test sites, participants, and users is easily performed via the CANTABTM administration platform. For training and general consultation, our experiences with the CANTABTM support staff have been unequivocally positive.
What are the next steps?
The next step for the AI-Mind team with respect to the cognitive data is curation. Despite the work it requires, one of the best aspects of CANTABTM is the large number of variables it outputs. Now, our team of neuropsychologists will begin the work of selecting and grouping variables to answer the various research questions of the project.
Can I enrol in the AI-MIND study?
Please see the AI-MIND website for details of how you can enrol in the study. To be eligible you must live in Norway, Italy, Finland or Spain, be 60-80 years old and be experiencing mild cognitive impairment or significant memory problems.
1. Rossini, P.M et al. “The Italian INCERCEPTOR Project”, Journal of Alzheimer’s Disease, vol. 72, no. 2, pp. 373-388, 2019, DOI: 10.3233/ JAD-190670
2. Hatlestad-Hall, Christoffer et al. ‘The organization of functional neurocognitive networks in focal epilepsy correlates with domain-specific cognitive performance’. Journal of Neuroscience Research, 26 Oct. 2021 : 2669-2687. http://dx.doi.org/10.1002/jnr.24896
3. Vecchio, Fabrizio et al. ‘Classification of Alzheimer’s Disease with Respect to Physiological Aging with Innovative EEG Biomarkers in a Machine Learning Implementation’. 1 Jan. 2020 : 1253 – 1261. http://dx.doi.org/10.3233/JAD-200171
The AI-Mind website
Oslo University Hospital
The AI-Mind project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 964220.
This article reflects only the author’s view and the European Commission in not responsible for any use that may be made of the information it contains.
Tags : cantab | cantab testimonial | alzheimer's disease | machine learning | cognitive impairment | mild cognitive impairment
Christoffer Hatlestad-Hall, Oslo University Hosptial