Brain Scan Model May Flag Alzheimer’s Risk Before Symptoms

Researchers developed a machine-learning model called ARC to assess how brain regions communicate, using resting-state functional MRI scans. They trained it with data from 180 adults over 75 known as SuperAgers, whose cognitive abilities remain unusually strong, and 110 people diagnosed with Alzheimer’s disease. The team also evaluated the model among 143 additional participants.

Next, ARC was used to assess 818 adults who had no signs of cognitive impairment and were somewhat younger than the SuperAgers. Although these adults performed similarly on memory and thinking assessments, the model placed them along a spectrum from SuperAger-like to Alzheimer’s-like brain activity.

Follow-up data spanning five to 15 years showed that participants with more Alzheimer’s-like patterns experienced faster cognitive decline, greater accumulation of amyloid-beta and a higher likelihood of developing clinical Alzheimer’s disease.

Lead author Kanhao “Alex” Zhao said the model’s predictive ability was comparable to or better than existing biomarkers, and that combining ARC with those measures improved prediction. The researchers said the approach could help identify elevated risk before noticeable memory problems and give scientists an earlier opportunity to study whether intervention can preserve cognitive abilities.

The study examined brain connectivity alongside established Alzheimer’s risk indicators, including genetic factors and proteins such as amyloid-beta and tau. The findings suggest the activity patterns may provide information beyond those biomarkers, though the research points to a potential tool for future study rather than an established clinical test.