Birkbeck researchers develop model for early detection of dementia

Dementia, a complex neurological condition affecting memory, cognitive function, and problem-solving abilities, often eludes early diagnosis. Traditional cognitive tests on their own, while essential, may not always provide a definitive diagnosis due to the influence of other health factors on cognitive function. As a result, other methods of detection are used alongside cognitive tests to determine if someone has dementia, such as genetic tests (looking at a person's genes), and biomarker tests (looking at specific substances in the body).  

The research was conducted by scientists from the Developmental Neurocognition Laboratory (DNL) at Birkbeck’s School of Psychological Sciences under the leadership of Michael Thomas, Professor of Cognitive Neuroscience. It incorporated intelligent modelling, specifically supervised machine learning models, to analyse vast datasets collected over different time intervals, in order to find out how effective different testing methods are at predicting whether someone will develop dementia in the future.  

Dr Samara Banno, lead researcher and Daphne Jackson Fellow in the School of Psychological Sciences, commented: “Dementia can be a heartbreaking condition that casts its shadow over every facet of a person's existence, and we are all potentially at risk. Once dementia starts, nothing yet can stop it. The best hope of treatment lies in interventions before dementia starts, and this is why our best defence lies in early detection. For it is in this early detection that we hold the power to shield our minds from its effects. The findings of this research may potentially revolutionise the diagnostic approach to dementia, offering a more comprehensive and accurate toolset.” 

The study involved monitoring individuals aged 60 to 75 over a four-year period. What emerged were key insights into predictive markers for early dementia diagnosis. Notably, certain genetic markers (such as the APOE e4 gene variant), specific biomarkers (such as tau protein in the brain), and measurements of the hippocampus (a brain region associated with memory), all demonstrated significant links to the likelihood of receiving a later dementia diagnosis.  

Perhaps most strikingly, the study suggested an association between brain structure measures, such as FDG (Fluorodeoxyglucose) uptake and hippocampus size, over the four-year monitoring period. The testing method of FDG positron emission tomography imaging, in particular, emerged as a promising tool for monitoring hippocampal function and structure in individuals with dementia, offering potential benefits for early diagnosis and treatment monitoring. 

A subsequent analysis of data from the team introduced cognitive and psychological scores alongside demographic data while excluding bio, genetic, and brain structure markers. These types of recall rely on a person’s ability to remember things that happened in the past, known as episodic memory, and to retrieve relevant words from language areas in the brain. 

Weaknesses in these skills, particularly at a young age such as in the forties and fifties, may be an earlier indicator of cognition decline even before mild symptoms of dementia appear – and a trigger for further clinical investigations. This shows the importance of using predictive models to prioritise different types of clinical risk markers and a sequence of investigation for earlier diagnosis of dementia.   

Ahead of its publication, Dr Banno and Professor Thomas will present Developing an Intelligent Prediction Model for Dementia at the Future Technologies Conference on 2 and 3 November, 2023.