Alzheimer’s Disease Rewires the Brain’s Cross-Hemisphere Dialogue, Study Finds

Alzheimer’s disease has long been understood as a disorder of memory and cognition, but the way it disrupts communication between the brain’s two hemispheres has remained frustratingly elusive. Now, a large-scale multicenter study published in BMC Medicine offers the most systematic picture yet of how this interhemispheric dialogue breaks down, revealing a striking anterior-posterior hierarchy of changes that could pave the way for a new neuroimaging biomarker of the disease.

The research, led by Chonggang Tong and Yuwei Su of the Beijing University of Posts and Telecommunications, together with colleagues, focused on homotopic functional connectivity, or HFC, the degree to which mirror-image regions in the left and right hemispheres synchronize their activity. This direct communication between corresponding brain areas is fundamental to integrating information across the two halves of the brain, supporting everything from motor coordination to language and attention. While previous studies had hinted at HFC alterations in Alzheimer’s disease, most were confined to small cohorts and simple statistical comparisons, producing inconsistent and often contradictory findings.

To overcome these limitations, the team assembled a dataset of 799 subjects drawn from seven different imaging sites, comprising 289 individuals with Alzheimer’s disease, 253 with mild cognitive impairment, and 257 cognitively normal controls. This scale matters. Multicenter designs introduce considerable variability in scanner hardware, acquisition protocols and participant demographics, but they also provide the statistical power and generalizability that small single-site studies lack. By pooling results across sites using random-effects meta-analysis, the researchers could distinguish robust disease effects from site-specific noise.

The team’s methodological pipeline began with functional magnetic resonance imaging, measuring blood-oxygen-level-dependent signals while participants rested in the scanner. From these data, they computed HFC for every participant, quantifying the temporal correlation between homotopic pairs of brain regions. They also derived a suite of structural asymmetry measures from structural MRI, including gray matter volume, fractal dimension, cortical thickness, sulcal depth and gyral index, allowing them to ask whether functional changes in cross-hemisphere communication track with structural differences between hemispheres.

For each imaging site, the researchers ran linear regression models with diagnosis as the main effect of interest and age and sex as covariates, then combined the resulting effect sizes across all seven sites through meta-analysis. The pattern that emerged was unambiguous and hierarchically organized. In posterior and subcortical regions, HFC was systematically reduced in Alzheimer’s patients. The reduction was particularly pronounced in the right hippocampus, with a Cohen’s d effect size of −0.541, and in the medial amygdala, with a Cohen’s d of −0.330, both surviving correction for false discovery rate. These regions, which include structures central to memory processing, showed the earliest and strongest erosion of interhemispheric synchronization.

In contrast, a focal and unexpected increase in HFC appeared in the prefrontal cortex, specifically in the medial anterior prefrontal area A9m, where the effect size reached a Cohen’s d of 0.391. This anterior-posterior gradient, with posterior and subcortical connectivity collapsing while certain prefrontal connections are strengthened, suggests that Alzheimer’s disease does not simply degrade cross-hemisphere communication uniformly. Instead, the brain appears to undergo a hierarchical reconfiguration, possibly reflecting compensatory mechanisms in frontal regions that attempt to offset deteriorating posterior networks, or perhaps a pathological redistribution of neural resources as the disease progresses.

Beyond mapping this pattern, the team tested whether HFC could serve as a practical diagnostic biomarker. They trained support vector machines, a widely used machine learning algorithm, to classify individuals as having Alzheimer’s disease or not, based on their HFC profiles. Crucially, they employed leave-one-site-out cross-validation, training the classifier on data from six sites and testing it on the held-out seventh, cycling through all seven configurations. This rigorous scheme ensures that the reported performance reflects genuine disease signals rather than scanner-specific artifacts. The classifier achieved a mean accuracy of 0.745 across sites, with an area under the curve of 0.815 and an F1 score of 0.774.

The researchers then asked whether adding structural asymmetry measures would sharpen the model’s diagnostic power. It did. Integrating gray matter volume, cortical thickness and the other structural metrics with HFC improved classification accuracy by up to 0.129, indicating that functional and structural signatures of hemispheric asymmetry carry complementary information about the disease. In a parallel analysis, support vector regression was applied to predict individual Mini-Mental State Examination scores, the standard bedside measure of cognitive status, from the same imaging features, probing whether the biomarker could capture disease severity rather than just diagnostic category.

The implications extend beyond the laboratory. Current biomarkers for Alzheimer’s disease, including cerebrospinal fluid assays and PET amyloid imaging, are expensive and invasive, while structural MRI alone captures late-stage neurodegeneration. A resting-state fMRI measure of homotopic connectivity, requiring no contrast agent and only a few minutes of scanning, could offer a cost-effective complement for early detection and for monitoring response to the growing arsenal of disease-modifying therapies. The finding that mild cognitive impairment sits between controls and Alzheimer’s disease in this hierarchy of connectivity changes further suggests HFC may track disease progression.

The study’s authors, whose work was supported by China’s National Science and Technology Major Project and the National Natural Science Foundation of China, caution that HFC is not yet a standalone diagnostic tool. Effect sizes at the individual level remain moderate, and clinical deployment will require validation in independent cohorts and harmonization of acquisition protocols. Nevertheless, the demonstration that interhemispheric dialogue is hierarchically reconfigured in Alzheimer’s disease, reproducibly across seven sites and nearly eight hundred brains, deepens understanding of the neural mechanisms underlying the disease and provides a concrete, measurable target for future therapeutic investigations and individualized assessment.

Subject of Research: Multicenter analysis of homotopic functional connectivity changes between brain hemispheres in Alzheimer's disease

Article Title: Hierarchical reconfiguration of interhemispheric dialogue in Alzheimer’s disease: a multicenter analysis

Article References: Tong, C., Su, Y., Zhang, W., Hu, P., Zhao, C., Liu, Y., & Zhong, S. (2026). Hierarchical reconfiguration of interhemispheric dialogue in Alzheimer’s disease: a multicenter analysis. BMC Medicine. https://doi.org/10.1186/s12916-026-05234-8

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05234-8

Keywords: Alzheimer's disease, homotopic functional connectivity, interhemispheric communication, neuroimaging biomarker, resting-state fMRI, mild cognitive impairment, support vector machine, multicenter study, brain asymmetry, hippocampus, prefrontal cortex, machine learning