Brain connectivity patterns predict Alzheimer’s years before memory loss

A Stanford Medicine-led study reveals that resting-state brain activity patterns can predict Alzheimer’s disease up to 12 years before cognitive symptoms or physical plaque accumulation appear, outperforming traditional genetic, tau, and amyloid-beta biomarkers

Published on November 4, 2026, in Science Advances, the team used resting-state functional magnetic resonance imaging (fMRI) to map neural firing networks in cognitively unimpaired adults, proving that specific functional connectome patterns can identify future neurodegeneration long before the physical buildup of amyloid-beta plaques or tau tangles.

Methodological design and the ARC scoring model

Traditional Alzheimer’s diagnostic research primarily benchmarks cognitively normal older adults against average peers, a method senior author Dr Yu Zhang notes is flawed because the “cognitively normal” baseline harbours wide neurobiological heterogeneity.

To overcome this limitation, the researchers contrasted two extreme biological cohorts aged 75 and older:

• Alzheimer’s cohort:

  • 110 adults with clinically diagnosed Alzheimer’s disease.

• Super-ager cohort:

  • 180 adults aged 75 and older who retained the memory performance and executive function of 45-year-olds.

By mapping the distinct regions of underactivity and overactivity separating these two extremes, the researchers established the Alzheimer’s resilience-vulnerability connectome (ARC) score, which differentiated super-agers from Alzheimer’s patients with 85% accuracy.

Longitudinal validation in mid-life adults

The researchers then validated the ARC score across a cohort of over 1,300 cognitively intact individuals aged 55 to 70:

Predictive superiority

Baseline ARC scores successfully stratified participants along a continuous spectrum of risk. Higher (Alzheimer’s-like) scores strongly correlated with accelerated memory loss, rapid executive function decline, and eventual conversion to Mild Cognitive Impairment (MCI) or Alzheimer’s within 5 to 12 years.

Outperforming conventional biomarkers

The ARC score demonstrated comparable or superior predictive accuracy for future cognitive decline compared to standard demographic data, genetic risk factors (such as APOE-ε4), and fluid/PET biomarkers for amyloid-beta and tau proteins. Combined models pairing ARC scores with existing biomarkers yielded the highest predictive precision.

Target for early therapeutic intervention

Unlike fixed genetic markers or late-stage protein plaques, functional connectivity signatures represent potentially modifiable neural circuits. Researchers plan to explore non-invasive interventions—such as targeted transcranial magnetic stimulation (TMS) and cognitive training—to actively guide vulnerable brain activity patterns toward the resilient “super-ager” profile before structural neurodegeneration occurs.