AI-Powered Screening Tool Helps Identify People at Higher Risk of Dementia

At the Alzheimer's Association International Conference (AAIC) 2026, researchers presented a new approach for identifying older adults who may be at increased risk of developing dementia. The study, titled "A Risk Stratification Pipeline to Enrich Study Populations for Early Dementia Detection in the PREDICTOM Project," describes an AI-assisted system designed to sort participants into higher- and lower-risk* groups before they undergo more expensive clinical testing.

The challenge facing researchers is straightforward: advanced assessments such as MRI scans, EEG testing, and biomarker analyses are costly and time-consuming. In a large study involving thousands of participants, it is simply not practical to perform every test on everyone.

To address this problem, the PREDICTOM team developed a risk stratification pipeline that combines two key ingredients: a dementia risk model based on established health and lifestyle factors, and a cognitive assessment model that evaluates whether an individual's thinking and memory performance is lower than expected for someone of the same age, sex, and educational background.

Unlike traditional screening approaches that rely on simple cut-off scores, the researchers aimed to distinguish between normal age-related changes and potentially concerning cognitive decline. A memory test score that may be entirely normal for a 75-year-old could be more worrying in a 55-year-old. The new cognitive performance model takes these differences into account.

One of the study's most notable findings was that , within the model development cohort, incorporating this age-adjusted cognitive model increased the identification of elevated-risk individuals by 28 percent compared with relying on risk factors alone. The researchers suggest that combining risk-factor information with personalized cognitive benchmarking may improve the efficiency of dementia screening programs.

The approach has already been put into practice within PREDICTOM. By May 2026, 1,180 participants had completed the study's first stage of home-based assessments and had been stratified according to their predicted risk. Researchers identified 189 higher-risk participants and 202 lower-risk participants, who can now be prioritized for more detailed follow-up investigations.

Importantly, the system is not designed to diagnose Alzheimer's disease. Instead, it functions more like an intelligent triage tool, helping researchers and clinicians decide who is most likely to benefit from additional testing. The ultimate goal is to create a scalable pathway in which people complete assessments at home, AI estimates their risk, and only those most likely to be at risk proceed to advanced investigations.

Researchers caution that the model has not yet been validated against definitive Alzheimer's biomarkers such as pTau217 blood tests or cerebrospinal fluid measurements. That work is planned as more data becomes available from later stages of the PREDICTOM study.

PREDICTOM offers a glimpse of how dementia screening could evolve in the coming years. Rather than sending every older adult for costly scans and specialist appointments, future healthcare systems may use AI-powered risk assessment to identify those most likely to benefit from further evaluation.

If successful, such an approach could help healthcare systems detect cognitive decline earlier, while making more efficient use of limited clinical resources.

*Note the lower-risk group is used as a reference to validate the approach in later stages of the project.

Click here to view posters presented at AAIC 2026 conference.