“An ARC Value Is Only a Snapshot, Not a Verdict,” Said Yu Zhang

“An ARC value is only a snapshot, not a verdict,” said Yu Zhang, principal investigator of a new Alzheimer’s study, this week.

Researchers at Stanford University have built an artificial-intelligence model that scores how the brain’s regions communicate while a person simply lies still in an MRI scanner. In a study published in Science Advances, that score flagged future Alzheimer’s disease up to twelve years before diagnosis, WELT reported.

The stakes are timing. By the time doctors diagnose dementia, the Stanford team writes, pathological changes in the brain have usually been running for years — and an analysis of 1,858 veterans, reported by ad-hoc-news.de, found that standard short screening tests often fail to capture the earliest deficits.

“Our model shows predictive power comparable to or better than established biomarkers,” said Kanhao Zhao, the study’s first author, “and that at an earlier point.”

MRI-Based Score Predicts Faster Memory Decline and Amyloid Buildup

  • The model reads functional connectivity — how brain regions talk to each other at rest — from a standard MRI scan. Nothing invasive, no injections.
  • People whose resting pattern sat near the Alzheimer’s end of the scale later showed faster memory decline and heavier amyloid buildup, even though their amyloid and tau values looked normal on the day of the first scan.
  • The score is not a test anyone can take yet. It remains a research tool, and its authors say so themselves.

A Score Built From Two Extremes

The Stanford team deliberately skipped the usual comparison group of cognitively normal older adults. Those controls are biologically too diverse, WELT’s report explains — some may already carry silent neurodegeneration that blurs the differences the model needs to see.

“An ARC Value Is Only a Snapshot, Not a Verdict,” Said Yu Zhang
Photo: ad-hoc-news.de

Instead, the researchers anchored the scale at its poles: Alzheimer’s patients on one end, and so-called super-agers on the other — people over 75 with the memory of a 45-year-old, estimated at roughly ten percent of older adults. The model was trained on 290 people, 180 super-agers and 110 Alzheimer’s patients, then checked against an independent group of 143, according to DIE WELT. It separated the two groups with about 80 percent accuracy.

How the Study Cohorts Compare

Cohort Group Key figure
ARC training set Super-agers / Alzheimer’s patients 180 / 110 of 290
ARC validation Independent group 143 people
ARC application Cognitively healthy adults, 55–70 More than 1,300
VA memory clinic Veterans, 2017–2025 1,858 records

Applied to more than 1,300 cognitively healthy adults between 55 and 70, the score did what the classic biomarkers could not. At the baseline scan, amyloid and tau — the misfolded protein deposits long suspected as Alzheimer’s triggers — did not differ between the groups at all. The resting communication pattern still predicted who would decline.

WELT’s report adds two layers. The score tracked the biological age of the brain rather than the number of birthdays, and it linked not just memory loss but apathy and depression to the risk pattern. First analyses point to two candidate genes, one governing inflammatory processes in cell metabolism and one building synapses, the connection points between neurons — candidates, the researchers caution, not confirmed hits, and larger studies must follow.

Tissue Loss a Decade First

That timing matters for treatment. Detecting changes in the first years of the disease process rather than after fifteen, the researchers argued, opens a window for intervention before symptoms begin.

“An ARC Value Is Only a Snapshot, Not a Verdict,” Said Yu Zhang
Photo: WEB.DE News

When Alzheimer Starts in the Legs

A third strand of reporting moves the story outside the brain entirely. Two Alzheimer's mutations produced deficits there anyway: PSEN1 A246E caused severe functional losses, the APP "Swedish" mutation moderate ones.

Two of the most common Alzheimer’s drugs, memantine and galantamine, did nothing for those peripheral deficits. Both are built to act on cognitive symptoms in the brain.

The clinical echo is visible in population data. In a UK Biobank analysis over twelve years, participants who later developed dementia already walked more slowly and gripped more weakly at the study’s start, and a clinical examination at Tokyo Medical University found muscle weakness and slowed gait in patients with early Alzheimer’s. Not every dementia is Alzheimer’s, WEB.DE News’s report cautions.

The three outlets thus draw three different clocks on the same disease. WELT’s account has function preceding the classic protein biomarkers by up to twelve years; the Johns Hopkins work has structure — tissue loss — leading symptoms by up to ten; WEB.DE News’s has the body signaling before the mind does.

None contradicts the others, but they measure different things on different scales, and only the Stanford score claims predictive power against established biomarkers.

When Memory Worries Warrant a Doctor

  • The ARC score is not clinically available. No scan you can currently obtain yields it, and a poor research result would say nothing definitive about any individual — Zhang’s own warning.
  • WEB.DE News’s report advises that anyone with a family history of early-onset Alzheimer’s raise it at preventive checkups.
  • Persistently slower walking, a weaker grip or balance problems, which the UK Biobank data tied to later dementia, warrant a clinical conversation when they appear alongside memory concerns.
  • Apathy and depression were linked to the risk pattern in the ARC data, WELT reported — another reason such changes should not be dismissed as mere mood.

A Trial to Shift the Pattern

The Stanford group’s next step is a therapy trial: testing whether targeted cognitive training can push a brain’s resting pattern toward the super-ager end of the scale, WELT reported. The gene candidates await confirmation in larger cohorts.

“The critical thing is that we intervene early,” Zhang said. “That’s why we focus on cognitively unimpaired people — to distinguish super-ager-like from Alzheimer-like brain patterns.”

References

  • Zhao, K. et al., Alzheimer’s Resilience-Vulnerability Connectome score — Science Advances
  • Neurology Open Access — Boston University / VA memory-clinic cohort analysis (1,858 veterans, 2017–2025)
  • NeuroImage Reports — Johns Hopkins University, presented at Neuroscience 2026

This article reports on peer-reviewed research findings and is not a substitute for individual medical advice. Concerns about memory, gait or mood should be discussed with a qualified physician.

Photo of author

Health Desk

Health Desk at Archyde.

Italy Coach Luca Banchi Names Squad for FIBA World Cup 2027 Qualifiers Against Turkey and Serbia

Slovakia Fire Chief Swears at Reporters Over Wife’s DUI, Then Apologizes