Digital Behavioral Testing and Early Indicators of Alzheimer’s Disease
Peer-reviewed studies examine AI analysis of brief clinical tasks for detecting amyloid and tau pathology
Two peer-reviewed studies from Linus Health and independent research teams, including investigators at Massachusetts General Hospital and the Harvard Aging Brain Study, report that brief digital tasks performed on a tablet can identify behavioral patterns associated with amyloid and tau pathology. Amyloid plaques and tau tangles are biological features linked to Alzheimer’s disease.
The studies analyzed how individuals draw, pause, and speak during a digital cognitive task lasting approximately three minutes. These behaviors were captured through sensors and processed using machine learning models. According to the findings, the resulting measures were associated with both cognitive impairment risk and biological markers of Alzheimer’s disease, including results from PET imaging.
One study examined data from nearly 1,000 participants and evaluated the ability of a digital clock and recall task to estimate cognitive impairment and amyloid positivity. The analysis found that a single assessment could support multiple predictions, a method described by the researchers as AI “multiplexing.” The results suggest that the task may help identify individuals who could benefit from additional testing, such as blood-based biomarkers, imaging, or specialist referral.
A second study followed more than 200 cognitively normal older adults over time. Participants with higher levels of amyloid or tau on PET scans showed greater change on a digital clock drawing task, particularly in measures related to processing speed and executive function. These changes were observed before differences were detected by standard cognitive tests.
Together, the studies indicate that digital analysis of everyday behaviors may detect early changes in brain function associated with Alzheimer’s pathology. The authors note that such tools could be integrated into primary care settings, where access to specialty cognitive evaluations is often limited, and used to support earlier clinical decision-making.
