Japanese researchers create AI models to screen for Alzheimer’s disease
By Carl Samson
Japanese researchers successfully developed machine learning models to predict elevated levels of brain amyloid beta (Aβ), a key indicator of Alzheimer’s disease risk.
Key findings: Three models demonstrated a significant ability to identify individuals with higher brain amyloid deposition. Using wearable sensors alone, the models showed a mean receiver operating characteristic — area under curve of 0.70, suggesting reliability. Integrating data from those sensors with demographic characteristics, as well as health and life environment questionnaire features, raised it to 0.79. According to the researchers, this is “fair and acceptable in real-world settings.” The AUC increased further to 0.83 with the addition of scores from the Japanese version of the Montreal Cognitive Assessment.
How the study was conducted: Researchers from Oita University and Eisai Co., Ltd., conducted the study from August 2015 to September 2019 in Usuki, Oita Prefecture. It started with 122 seniors with mild cognitive impairment or subjective memory complaints. Participants wore the sensors to track lifestyle factors and completed self-reported questionnaires, cognitive test and positron emission tomography (PET) imaging. After the third year, the study had 282 records eligible for analysis.
Why this matters: The study presents a non-invasive, cost-effective method to identify an individual’s risk for Alzheimer’s disease, which makes up 60 to 70% of all dementia cases, according to the World Health Organization. On top of cognitive and blood tests, current screening methods include lumbar puncture — which involves extracting cerebrospinal fluid from the lower back — PET, magnetic resonance imaging and computed tomography.
What’s Next: The models will need to be refined for broader application and integration into clinical practice. For now, clinicians may be able to use them to help decide which patients need more invasive diagnostic procedures.
The study was published in the journal Alzheimer’s Research & Therapy on Dec. 12, 2023.
Share this Article
Share this Article
