Objectives

This study investigates the value of a multi-parametric [18F]PI-2620 tau PET/MRI protocol to distinguish different Alzheimer's disease (AD) variants: amnestic AD (aAD), logopenic variant primary progressive aphasia (lvPPA), and posterior cortical atrophy (PCA).
It demonstrates that hybrid PET/MRI with [18F]PI-2620 provides phenotype-specific information about tau accumulation, brain atrophy, brain perfusion, functional network alterations, and white matter microstructural alterations.
Refinements in the Braak staging system are needed for the non-aAD cases, particularly lvPPA and PCA.

Methodology

Dynamic PET data were acquired and reconstructed using an ordered subset expectation maximization algorithm.
Kinetic modeling was performed using the Multilinear Reference Tissue Model 2 (MRTM2) with cerebellar cortex (excluding the dentate nucleus) as the reference region to generate DVR and R1 parametric images.
Voxel-based morphometry (VBM) was performed using the computational anatomy toolbox (CAT12) in SPM12.
rs-fMRI data were pre-processed using SPM12, including slice-time correction, realignment, normalization, and smoothing. Functional connectivity of the posterior cingulate cortex (PCC) was computed.
Diffusion-weighted MR data were processed to generate fractional anisotropy (FA) maps.
Voxel-wise differences between groups were tested using two-sampled t-tests within SPM12. VOI-based statistical analysis was performed using different VOI sets for DVR, R1, and GMD maps.

aAD patients showed significantly higher tau accumulation (DVR) in the temporal lobe, anterior cingulate cortex, and anterior prefrontal cortex compared to PCA patients (p < 0.001).
PCA patients showed significantly higher DVRs in the secondary visual cortex and visual associative cortex compared to aAD patients (p < 0.001).
lvPPA patients showed significantly higher DVRs in the vermis of the cerebellum compared to aAD and PCA patients (p < 0.001).
R1 values were lower in the occipital lobe in PCA patients compared to aAD and lvPPA patients (p < 0.001).
VBM analysis showed decreased GMD in the occipital lobe in PCA patients compared to lvPPA patients, and different GMD patterns between lvPPA and aAD (p < 0.001).
Functional connectivity of the PCC was reduced in PCA and aAD patients compared to lvPPA patients, in specific brain regions (p<0.005, cluster size > 30).
Bilateral cingulum had lower FA values in lvPPA compared to aAD (p=0.029), primarily driven by the left cingulum (p = 0.029).

The study's sample size, particularly for the lvPPA and PCA groups, is relatively small, limiting the statistical power and generalizability of the findings. Future studies should aim for larger, more balanced cohorts.
The study lacks detailed neuropsychological assessments beyond MMSE/MoCA. Including comprehensive cognitive and behavioral data would strengthen the phenotypic characterization and allow for correlations with imaging findings.
The choice of the cerebellar cortex as a reference region for [18F]PI-2620 analysis has been recently questioned. While the authors excluded the dentate nucleus, the potential impact of using an alternative reference region (e.g., fusiform gyrus) should be discussed.
The study relies on cross-sectional data. Longitudinal data would be valuable to track the progression of tau pathology and other imaging markers over time in different AD variants.
While the study mentions the cascading network failure hypothesis, a more in-depth discussion of how the findings specifically support or challenge this theory would be beneficial.