Tau and Thickness SVD Loadings
Spatial representation of the tau and thickness loadings for the first canonical component. The strongest positive weights in the tau eigenimage are in areas of the entorhinal cortex and fusiform gyri, while the strongest negative weighs corresponding to the cortical thickness map appear within the entorhinal cortex and posterior cingulate cortex.
SVD Scores
Tau and thickness individual scores for the first SVD canonical component in the latent space. The tau and thickness scores are highly correlated, and the correlation strength does not depend on the APOE ε4 genotype status. In contrast, the thickness and amyloid scores appear to show a weaker association.
An SVD analysis was applied to a matched cross-sectional sample of cortical thickness maps and Tau PET images. The first canonical component accounted for 21.56% of the total co-variability between thickness and tau.
The strongest positive weights in the first tau eigenimage correspond to the entorhinal cortex and fusiform gyri. Such regions appeared to be negatively correlated with the strongest negative loads in the cortical thickness eigenimage, which correspond to the entorhinal cortex and posterior cingulate cortex.
The scatter plot for the first SVD-based tau and thickness scores shows a high correlation, particularly for the APOE ε4 subjects. This data is a compelling evidence of an overall linear dependency between thickness and tau.
In contrast, the first amyloid scores showed lower correlation with the thickness scores. The strength and direction of this linear association did not change when the data was segregated by the APOE ε4 status.
Taken together, these linear associations show that tau has a greater implication than β-amyloid in the process of cortical thickness reduction due to neurodegeneration in MCI.