Cambridge scientists have created an intelligent tool that can predict in 4 out of 5 cases if individuals showing early dementia symptoms will progress to Alzheimer’s disease.
This method could potentially reduce the need for invasive and costly diagnostic procedures, enhancing treatment outcomes with timely interventions like lifestyle changes or appropriate medications.
Dementia is a significant global health challenge, affecting over 55 million people worldwide and costing approximately $820 billion annually. The number of cases is projected to almost triple over the next 50 years.
Alzheimer’s disease is the primary cause of dementia, contributing to 60-80% of cases. Early detection is crucial for effective treatments, but current diagnostic tools relying on invasive or expensive tests like PET scans or lumbar punctures may not always be available at memory clinics. This leads to possible misdiagnoses and delays in effective treatment for up to a third of patients.
A team from the University of Cambridge’s Department of Psychology has developed a machine learning model that can forecast the progression of mild memory and cognitive issues to Alzheimer’s disease. Their study, published in eClinical Medicine, demonstrates its superior accuracy compared to existing clinical diagnostic methods.
Using non-invasive and low-cost patient data, including cognitive tests and MRI scans displaying grey matter atrophy from a pool of over 400 volunteers in the USA, the researchers established their predictive model.
Validated with real-world patient data from another 600 participants and longitudinal information from 900 individuals in memory clinics in the UK and Singapore, the algorithm accurately distinguished between those with stable mild cognitive impairments and those progressing to Alzheimer’s within three years.
This model showed an 82% accuracy in identifying individuals developing Alzheimer’s and an 81% accuracy in predicting those who did not, solely based on cognitive tests and MRI scans.
Compared to standard clinical markers, the algorithm was three times more precise in predicting Alzheimer’s progression, emphasizing its potential to reduce misdiagnoses.
Furthermore, the model categorized individuals with Alzheimer’s at their initial memory clinic visit into three groups based on their expected progression speed, aiding in identifying those who may benefit from early treatments and those who require close monitoring.
These advancements can guide individuals with stable symptoms toward appropriate care pathways, distinguishing them from those with conditions like anxiety or depression.
The researchers aim to expand their model to other forms of dementia and utilize various data sources like blood test markers to enhance its applicability.
