Eisai harnesses wearables data for AI-led Alzheimer's prediction

Its AI model can predict the accumulation of amyloid-beta protein, a major Alzheimer's biomarker.
By Adam Ang
04:02 am

Photo: MStudioImages/Getty Images

Japanese pharmaceutical company Eisai, together with Oita University in Japan, has built what could be the world's first AI model that uses data from wearable devices to predict Alzheimer's disease, the most common form of dementia.

Their research team embarked on a study to create a cost-effective, practical tool to pre-screen people suspected of developing the disease. Their findings were published in the journal Alzheimer's Research & Therapy.


The study collected biological and lifestyle data from 122 individuals aged 65 and older with mild cognitive impairment or subjective memory impairment. Biological data, including physical activity, sleep, and heart rate, were collected from wristband sensors they were made to wear for seven days every three months from 2015 to 2019. The following lifestyle data were also collected from their medical consultations: employment status, frequency of going outdoors, means of transportation, number of days participating in community activity, and backgrounds (age, education history, history of alcohol use, and medical history). 

Moreover, participants underwent annual amyloid PET (positron emission tomography) examinations, which detect the accumulation of amyloid-beta protein in the brain – a crucial biomarker of Alzheimer's. 

Researchers developed a predictive model that combined three machine learning technologies: support vector machine, Elastic Net, and logistic regression, to integrate all the collected data. Its purpose is to determine how likely each participant will test positive by amyloid PET screening.

An evaluation of the AI model showed that it has a "sufficient" capability to predict amyloid-beta protein accumulation. Through this model, the study was able to identify 22 factors that contribute to amyloid-beta protein accumulation: physical activity, sleep, heart rate, amount of conversation, age, length of education, living with or without children, means of transportation, presence of accompanying person for hospital visits, communication frequencies, and number of outings.


Progress in Alzheimer's research led to the discovery of amyloid-beta protein accumulation in the brain as a significant biomarker of Alzheimer's. This further resulted in the development of therapeutic drugs targeting it.

Still, for incurable diseases such as Alzheimer's, screening is imperative. Currently, testing is possible through amyloid PET and cerebrospinal fluid (CSF) testing, which are costly, limited, and invasive, according to Eisai and Oita University researchers. 

AI has been applied to explore ways to make Alzheimer's testing accessible for many, especially now that countries like Japan are ageing at a faster pace. While there are already studies that have applied AI to predict amyloid-beta protein accumulation, these have so far only used cognitive function tests, blood tests, and brain imaging tests. The latest research by the Eisai and Oita University team provides value to this growing area of study as it focuses on lifestyle and biological data. 

Following their pioneering work, the team sees a potential for wider application of their AI model to pre-screen people for Alzheimer's, particularly those with little access to amyloid PET and CSF testing. 


The Japanese research team was not the first to contemplate using wearable devices to track or predict biomarkers of Alzheimer's. In 2020, Alzheimer’s Research UK announced a project to develop a wearable diagnostic device for the early detection of neurodegenerative diseases with funding from Microsoft founder Bill Gates. 

In recent years, more projects have been carried out to improve Alzheimer's detection by leveraging AI technology. Researchers from Australia's CSIRO and Queensland University of Technology recently came out with an AI-based benchmark for measuring brain atrophy, touted to be a global first. Fujifilm also developed an AI tool that was found to accurately predict patients who would progress to Alzheimer's within two years. Lotte Healthcare from South Korea recently announced a partnership with iMediSync to explore developing new AI-driven healthcare services, including screening for neuropsychiatric disorders. 

Meanwhile, Eisai continues its involvement in Alzheimer's research. Recently, it launched a new digital health business called Theoria, which has a specific focus on dementia. Launching in April, Theoria is coming out with its flagship solution, a risk prediction algorithm for the early detection of mild cognitive impairment.