Spatially distributed tau and amyloid PET scores have been shown to correlate with glucose metabolism in Mild Cognitive Impairment (MCI) independently of APOE ε4 genotype (Carbonell, 2020; Carbonell, 2025). However, the relationship between tau, β-amyloid, and brain atrophy remains unclear.
Glucose metabolism, measured with FDG-PET, reflects functional impairment, while cortical thickness, cortical volume, and gray matter (GM) density indicate structural neurodegeneration. Although both biomarker types are associated with Alzheimer’s disease (AD), they capture distinct aspects of disease progression. Atrophy biomarkers, such as reductions in cortical thickness, volume, and GM density, are well-established indicators of neurodegeneration, correlating with neuronal loss, synaptic degeneration, and cognitive decline.
We hypothesize that tau PET scores are more strongly associated with atrophy measures than with glucose metabolism, independent of β-amyloid and APOE ε4 effects. To investigate this, we apply Singular Value Decomposition (SVD), a multivariate technique that identifies uncorrelated components in the spatial organization of linear associations.
Schematic illustrating the hypothesis that Tau PET is more strongly associated with cortical atrophy measures than β-amyloid.
References
Carbonell et al., J. Alzheimers Dis., 73: 543-557, 2020; doi.org/10.3233/JAD-190560
Carbonell et al., Alzheimers. Dement., 21: e14625, 2025; doi.org/10.1002/alz.14625
It has been shown that spatially distributed scores from Tau and Amyloid PET images are significantly correlated with glucose metabolism in Mild Cognitive Impairment (MCI) independently of the APOE ε4 genotype.
In contrast, the relationship between tau, β-amyloid and atrophy has not been fully understood. Glucose metabolism measured with FDG PET reflects functional impairment in the brain, specifically reduced neuronal activity and metabolism, while cortical thickness, cortical volume, and gray matter (GM) density measure structural damage due to neurodegeneration. That is, while both type of biomarkers are associated with Alzheimer’s disease, they capture different dimensions of the disease. While brain atrophy can be estimated with several measures, the decrease in cortical thickness, cortical volume, and GM density are well-established atrophy biomarkers and have been associated with neurodegeneration in Alzheimer's disease. The loss of neurons and synapses leads to this decrease, which corresponds to cognitive decline and can be used to track disease progression.
We hypothesize that spatially distributed scores of Tau PET have a stronger association with atrophy measures than with glucose metabolism. Also, such a dependency occurs independently of the effects of β-amyloid and APOE ε4 genotype. With this aim, we employed the multivariate technique of Singular Value Decomposition (SVD). The SVD approach produces uncorrelated, canonical correlation components in the spatial organization of the linear association between two different datasets.
We found that the SVD-based tau scores showed extended areas of statistical significance with cortical thickness. Although less spatially extended, the cortical volume and the relative and absolute GM densities also showed areas of tau-related reductions. More importantly, such spatial relationships appeared to be stronger than the corresponding Tau-FDG association and occurred independently of the concurrent effects of β-amyloid and the APOE ε4 genotype.