An fMRI marker of Alzheimer’s-related cognitive decline

Researchers at Université Savoie Mont Blanc, Radboud University and the University of Oxford’s Wellcome Center for Integrative Neuroimaging, report that attenuation of the brain’s intrinsic anticorrelation between the default mode and dorsal attention networks serves as an fMRI biomarker of Alzheimer’s disease-related cognitive decline that is partly independent of tau pathology and education-based cognitive reserve.

Alzheimer’s disease still lacks a single reliable biomarker that can pinpoint when and where abnormal protein accumulation leads to network failure and irreversible structural damage. Current understanding of large-scale functional breakdown in Alzheimer’s disease remains incomplete, and proposed mechanisms connect amyloid and tau pathology, vascular brain lesions, and neurodegeneration.

Beta-amyloid accumulation begins many years before clinical symptoms and supports a view of Alzheimer’s disease as a continuum from normal aging to dementia rather than a discrete category, pushing researchers to look beyond accumulation for insights.

Dueling networks

Neurocognitive networks in the human brain can show patterns of cooperation or competition, and resting-state functional MRI has been used to study how those networks interact without requiring a task.

The default mode and dorsal attention networks are two large-scale brain systems that typically have an on/off relationship where, when one becomes active, the other deactivates. If one’s mind is wandering through daydreaming or moments of self-reflection, remembering yesterday’s lunch, or pondering a future vacation, it is the default mode in operation.

If, instead, one is trying to hit a baseball, make a precision cut to a piece of wood, or editing a science writer’s occasionally irredeemable grammatical mistakes before publication, it is the dorsal attention keeping the mind on task.

Previous studies reported that resting-state functional connectivity within and between large-scale networks supports cognitive performance in healthy adults and that Alzheimer’s disease involves alterations in these connectivity patterns linked to cognitive decline. A “functional imbalance” in resting-state networks has also been proposed more generally as a potential basis for neurodegenerative pathology.

Resting-state connectivity, including the interaction between the default mode network (DMN) and dorsal attention network (DAN), could provide potential systems-level signatures related to pathology burden, cognitive decline, and concepts such as cognitive reserve.

Probing default and attention networks

In the study, “The intrinsic connectivity between the default mode and dorsal attention networks is an independent fMRI biomarker of Alzheimer’s disease pathology burden,” published in NeuroImage, researchers analyzed MRI, PET amyloid and tau imaging, and cognitive and demographic data to see whether DMN–DAN anticorrelation associates with AD pathology burden (amyloid + tau) and to test whether it represents a mechanism independent of tau-driven spread; you capture this reasonably.

Final analyzed data came from 182 participants in the ADNI 3 cohort with resting-state fMRI and structural MRI at a baseline session, along with cognitive evaluations and demographic information. Resting brain activity was parsed into large-scale networks such as the default mode and dorsal attention networks, then statistical tools measured how strongly those networks rose and fell together over time.

Patterns of this network coupling were compared across different levels of amyloid load and cognitive performance to test whether the strength of the default-attention anticorrelation tracked Alzheimer’s-type changes.

From connectivity to cognition

A reduced separation between default mode and dorsal attention signals appeared in participants with high amyloid burden and cognitive impairment, making these networks less distinct in their activity. Without the clear on/off separation, thinking becomes more vulnerable to interference, making it harder to stay focused, retrieve information, and coordinate complex tasks.

In the full cohort, weaker on/off anticorrelation between default mode and dorsal attention networks went hand in hand with lower cognitive scores. Multivariate models that included age, sex, education, and tau burden in a temporal cortex meta-ROI still left room for DMN-DAN connectivity as an independent predictor, with roughly 5% of the variance in cognitive performance linked to this single fMRI measure.

Analyses extended the relationship to Mini-Mental State Examination scores and to domain-specific tests drawn from the ADNI neuropsychological battery, again tying variability in DMN-DAN anticorrelation to cognitive performance across multiple domains. Education level, used as a stand-in for cognitive reserve, did not shift that association.

Rethinking mechanisms of decline

Findings from this work place a simple resting-state measure at the center of a complicated disease story. Network separation that usually helps keep inward and outward focus distinct appeared least intact in amyloid positive, cognitively impaired participants.

Rather than pointing to a single cascade starting from one structure, the authors described a multifocal pattern of “strikes” across cerebral and extracerebral systems, including vascular disease, sleep loss, and stress that ultimately converge on a clinical syndrome.

Current results differ from interpretations that place DMN-DAN changes mainly as intermediaries of tau-dependent neurodegeneration. Network behavior in this cohort acts less like a readout of clinical labels and more like a candidate signal that travels alongside pathology.

Patterns suggest that this coupling measure could serve as a functional marker that captures a mechanism of cognitive dysfunction not fully explained by advanced temporal tau or by cognitive reserve.

Future research directions described in the paper include longitudinal and interventional studies to clarify whether network dysfunction precedes or follows abnormal protein accumulation, as well as development of personalized multicause risk models that incorporate normative DMN-DAN connectivity patterns, pathology measures, and lifestyle factors across large longitudinal datasets.

© 2025 Science X Network