Singh‐Reilly, Neha

Background Frontotemporal dementia is an umbrella term that encompasses several clinical syndromes with impaired behavioral, language, and motor functions. These syndromes show considerable overlap in clinical features and imaging patterns. Therefore, there is a need to investigate the syndromic heterogeneity in FTD using unbiased data-driven approaches. Methods We used data-driven clustering analysis of structural magnetic resonance imaging (MRI) data on 400 patients with clinical FTD diagnoses [behavioral variant of frontotemporal dementia (bvFTD), semantic variant of primary progressive aphasia (svPPA), right temporal variant of frontotemporal dementia (rtvFTD), apraxia of speech with agrammatic aphasia (AOS-PAA), primary progressive apraxia of speech (PPAOS), progressive supranuclear palsy (PSP), corticobasal syndrome (CBS) and primary progressive aphasia who did not fit into the other diagnostic categories (PPA-other)]. MR images were w-scored relative to cognitively unimpaired individuals, and principal component analysis was performed. A clustering ensemble approach, including hierarchical algorithms, was applied to the MR-based principal components, and imaging and clinical characteristics of the clusters were investigated. Various numbers of clusters ( K = 2, 3, or 4) were evaluated. Results The K = 3 solution offered the most clinically meaningful separation of FTD syndromes. The first cluster captured mostly frontal MRI abnormalities related to the speech, language and behavioral clinical dimensions, including patients with AOS-PAA, PPAOS, PPA-other, and bvFTD. The second cluster captured mostly temporal abnormalities and included mainly patients with svPPA and rtvFTD, but also bvFTD, AOS-PAA, and PPA-other. The third cluster captured cortical and subcortical atrophy, particularly in the midbrain, and included atypical Parkinsonian syndromes, with all PSP and CBS patients captured in this cluster, as well as most PPAOS patients. Considerable overlap of clinical syndromes was noted across these clusters, whereby patients with AOS-PAA, svPPA, PPA-other, and bvFTD were captured in more than one cluster. Discussion Our findings highlight heterogeneity in FTD, which mainly exists along three axes: speech, language and behavioral deficits reflecting frontal atrophy, language deficits reflecting temporal atrophy, and motor and motor speech deficits reflecting mostly midbrain and subcortical atrophy, with cortical involvement.

Abstract Background Atypical clinical presentations of Alzheimer's disease (AD) are associated with heterogeneous patterns of cortical tau‐PET uptake. The medial‐temporal (MTL) meta‐ROI is typically used to determine tau‐PET abnormality in AD, but this approach may not detect focal uptake in atypical patients. Our goal was to develop regional tau‐PET cut‐points and determine whether the number of positive regions better captures positive tau‐PET signal in atypical AD. Method 127 amyloid‐positive atypical AD (visual=68, language=39, motor=15, others=5) and 62 amyloid‐negative cognitively unimpaired (CU) individuals were recruited by the Neurodegenerative Research Group, Mayo Clinic, and underwent, amyloid‐PET and 18 F‐AV‐1451(tau)‐PET. Images were parcellated using the modified Harvard‐Oxford atlas and we focused on 67 AD‐specific regions. Rather than using the average SUVR across voxels within a region to quantify uptake, we used the 95 th percentile of SUVR values across voxels within a region to capture focal cases. We then used receiver operating characteristic (ROC) methods and a Youden threshold to define regional abnormality. The number of abnormal regions was calculated for each participant and used to discriminate between CU and atypical AD at the individual level. Result Across 64 regions, our approach showed excellent (AUROC>0.90) differentiation between atypical AD and CU. The temporal‐fusiform, parahippocampal gyrus and inferior temporal gyrus showed the best differentiation (Figure 1). All 67 regions showed excellent specificity (>85%), with 63 regions showing excellent sensitivity (≥80%). The lateral occipital cortex, temporal‐fusiform and parahippocampal gyrus showed the best specificity and sensitivity (Figure 2). The distribution of the number of tau‐elevated regions is shown in Figure 3. 74% of CU had no positive regions, while 43% of the atypical AD patients had all 67 regions affected. An optimal threshold of 7 tau‐elevated regions classified all CU below, and 123 (97%) atypical AD above threshold. Using a standard approach based on the average SUVR in the MTL meta‐ROI, only 116 (91%) atypical AD patients were tau‐PET positive. Conclusion Tau uptake in temporal regions showed the greatest group‐level differentiation. However, examining regional tau‐positivity may offer better sensitivity in atypical AD patients compared to the temporal meta‐ROI, likely due to the presence of focal non‐temporal patterns of uptake.

Abstract Background Quantitative susceptibility mapping (QSM) can detect iron deposition or myelin loss in the brain by estimating magnetic susceptibility properties. Iron is an important component in neurofibrillary tangles and is known to co‐localize in regions with tangle burden in Alzheimer's disease (AD). Abnormal magnetic susceptibility measured on QSM has been observed in atypical AD, although it is unclear if susceptibility is regionally related to AV‐1451‐PET uptake in atypical AD. Method Forty atypical AD patients (visual variant=24, language variant=7, others=9) were recruited by the Neurodegenerative Research group from the Department of Neurology, Mayo Clinic, Rochester, MN. They underwent a 3T MRI scan with a five‐echo gradient echo sequence for calculation of QSM ( n = 80), Aβ (Pittsburgh Compound‐B) and a 18 F‐ AV‐1451 PET scan ( n = 73). All patients showed evidence of Aβ positivity on PET. To assess the relationship between QSM and AV‐1451‐PET, two approaches were used i) Voxel‐based regression analysis using VoxelStats for the whole brain, and ii) region‐of‐interest (ROI)‐based Spearman's correlation analyses using cortical and subcortical ROIs. All models were adjusted for age and sex as covariates and were corrected for multiple comparisons, specifically the VoxelStats regression maps were corrected using random field theory and the ROI‐based correlation plots were corrected using false discovery rate. Result At the voxel‐level, positive correlations between AV‐1451‐PET and QSM were only observed in the left caudate (Figure 1). At the ROI‐level, no significant associations were found, although uncorrected data showed a positive association in the occipital lobe (red circle) and a negative association between substantia nigra susceptibility and occipital AV‐1451 uptake (blue circle) (Figure 2). Since these findings did not survive corrections, they should be interpreted cautiously. Conclusion Our data provides little evidence that regional AV‐1451 PET uptake is related to susceptibility changes in patients with atypical AD, suggesting that iron deposition may not be associated with regional tau accumulation in AD.

Abstract Background In addition to visuospatial and visuoperceptual deficits, most PCA patients also report deficits in visual attention, episodic memory, and working memory. The dorsal attention network is known to play a critical role in modulating these clinical functions. However, little is known about the relationship of the dorsal attention network with other core networks and relationships to volume loss and memory functions in PCA. Method Fifty‐seven PCA patients were recruited by the Neurodegenerative Research group, Mayo Clinic, and compared to 60 cognitively unimpaired (CU) individuals from the Mayo Clinic Study of Aging. The CONN functional connectivity toolbox was used to calculate within‐ and between‐network connectivity. Functional connectivity within the key regions (i.e., inter‐hemisphere connectivity) of the dorsal attention network, namely frontal eye field (FEF) and intraparietal sulcus (IPS), and the entire network was generated. Functional connectivity between the dorsal attention network and the frontoparietal, visual, and default mode networks (DMN) were also generated using CONN's network parcellation atlas. Multivariate linear regression models were fit i) to compare within‐ and between‐network connectivity between both groups and ii) to assess relationships between within‐network connectivity and both gray matter volumes and clinical test scores. Result PCA showed reduced within‐network connectivity in the dorsal attention network, specifically within the IPS and the entire network compared to CU. An increase in between‐network connectivity between the frontoparietal network and both FEF and the entire dorsal attention network was noted in PCA compared to CU (Figure 1). Lower dorsal attention within‐network connectivity was associated with a trend for lower volumes in the entire network (Figure 2) and significantly lower scores on the auditory verbal learning test‐recognition percent correct (AVLT‐RCP) and Wechsler Memory Scale (WMS) III‐digit span backward in PCA (Figure 3). Conclusion These findings suggest that within‐ and between‐network connectivity of the dorsal attention network, both within key regions and the entire network, was disrupted in PCA patients. However, this network did not show disrupted connectivity with the core PCA networks. Within‐network connectivity disruptions in the dorsal attention network are associated with aspects of memory function, and to some extent, with volume loss.