Posted on 26 October 2020
Longevity briefs provides a short summary of a novel research, medicine, or technology that caught the attention of our researchers in Oxford, due to its potential to improve our health, wellbeing, and longevity.
Why is this research important: While there is currently no cure for Alzheimer’s disease, early detection can help to delay and treat the condition to improve outcomes. As such, an automated test to predict the onset of Alzheimer’s disease could be highly beneficial.
What did the researchers do: Researchers from IBM and Pfizer collected descriptions of the ‘cookie theft’ scene (pictured below) written by elderly participants of the Framingham study. They then trained an AI model to recognise language impairment in these descriptions, which is often used to test cognitive function, and studied whether this model could accurately predict whether participants would develop Alzheimer’s disease.
Key takeaway(s) from this research: The model was able to predict the clinical diagnosis of Alzheimer’s disease by age 85 with 71% accuracy. On average, this prediction was made 7 years before the disease was diagnosed. Comparatively, predictions based on biomedical data (such as age, gender, and conditions like diabetes and hypertension) achieve around 59% accuracy. This suggests that language performance contains subtle warning signs for Alzheimer‘s prior to clinically detectable cognitive impairment.
Researchers hope to further improve the model by combining it with other data such as geographic and socioeconomic information.