Citations

AbstractBackgroundDisproportionately enlarged subarachnoid‐space hydrocephalus (DESH), characterized by tight high convexity CSF spaces, ventriculomegaly, and enlarged Sylvian fissures, is thought to be an indirect marker of a CSF dynamics disorder. Although this imaging phenotype was originally described in normal pressure hydrocephalus (NPH), it has since been more widely recognized in the aging population with a prevalence of 6‐7% in the Mayo Clinic Study of Aging (MCSA) 1. The clinical significance of DESH with regards to cognitive decline outside of NPH is not yet well‐defined. The goal of this work is to determine if DESH is associated with risk of cognitive impairment.MethodParticipants in the population‐based MCSA who met the following criteria were included: age ≥ 65 years, 3T MRI, cognitively unimpaired at enrollment, and at least one follow‐up visit. A support vector machine based method to detect the DESH imaging features on T1‐weighted MRI was used to calculate “computational DESH” or CDESH scores2 (Figure 1). Cox proportional hazard models were fit with events defined as progression from cognitively unimpaired to mild cognitive impairment or dementia. Time was defined as years from MCSA enrollment to the first visit with cognitive impairment with censoring at last follow‐up. CDESH was modelled as a spline to allow for non‐linear effects. Covariates included age, sex, education, APOE status, cortical thickness, white matter hyperintensity (WMH) volume, and total intracranial volume.ResultParticipant demographics are summarized in Table 1. The model including the above co‐variates showed a significant CDESH effect (p=0.02) with an approximately 20% greater hazard of progression for a CDESH score of +1 versus −1 (HR 1.2, 95% CI 1.0‐1.4) (Figure 2).ConclusionImaging features of disordered CSF dynamics are an independent predictor of subsequent cognitive decline in the MCSA, among other well‐known factors including age, cortical thickness, and APOE status. Therefore, since DESH contributes to cognitive decline and is present in the general population, identifying individuals with DESH features may be important clinically and also used as a selection stratification variable for clinical trials. References: (1) Graff‐Radford et al doi:10.1212/WNL.0000000000008616; (2) N. Gunter et al doi:10.1016/j.nicl.2018.11.015.

AbstractBackgroundWe report the associations between change in renal biomarkers in CKD and non‐ CKD participants and change in structural MRI outcomes over 3 years in the BRain IN Kidney disease (BRINK) study MRI cohort. The imaging outcomes were cortical thickness in Alzheimer’s disease signature regions, white matter hyperintensities (WMH), and fractional anisotropy in the corpus callosum (FA CC), an early marker of cerebrovascular disease.MethodsParticipants were ≥ 45 years with CKD (estimated glomerular filtration rate (eGFR) in mL/min/1.73 m2< 60; non‐ dialysis) or non – CKD (eGFR ≥ 60) in Minneapolis/St. Paul. CKD severity was measured by the eGFR and urine albumin creatinine ratio (UACR‐albuminuria, a measure of microvascular disease). Brain images were obtained on a 1.5 T Phillips Ingenia machine at baseline and 3 years. Linear regression models measured the associations between 3 year change in eGFR and UACR, as predictors of change in MRI measures: AD cortical thickness, WMH, and FA CC over 3 years. Adjusted models included age, gender, AA race, education years, cerebrovascular disease risk factor score (CVD RF), smoker/alcohol use, and pulse pressure as covariates, and TIV for change in WMH models.ResultsAt baseline, mean age of the 97 participants was 67 years, and mean eGFR was 37 for the CKD group vs. 83.2 for non‐ CKD (Table 1). Mean change in eGFR and UACR in the CKD group were ‐3.6 and +212, respectively compared with ‐ 8.5 and +1.7 in the non‐ CKD group. Mean WMH volume increase (+5.8 cc) and mean FA CC decrease (‐0.012) were significant in CKD (p < 0.001) for both (Table 2). In adjusted models both a 1% decrease and a ≥ 3 unit/year decline in eGFR were associated with a decline in FA CC over 3 years in the CKD group (β (SE) =‐0.020 (0.007); p = 0.004) and ‐0.0097 (0.0042); p = 0.025, respectively). Cortical thickness models were not significant. Unexpectedly, change in UACR was not associated with change in any MRI outcomes.ConclusionsDecreased eGFR was associated with decreased FA CC in CKD, suggesting that CKD progression exerts its influence largely through cerebrovascular mechanisms.

AbstractBackgroundThe NIA‐AA Research Framework on Alzheimer’s disease (AD) represents an important advance in the biological characterization of the AD spectrum. While this framework predicts disease progression, definitions of amyloid, tau, and neurodegeneration positivity based on cut‐scores may limit the performance of predictive models.Method252 cognitively normal (CN), 424 amnestic mild cognitive impairment (aMCI) and 128 AD participants were included in this study. aMCI participants at baseline were classified using 2 different methods: 1) Using conventional AT[N] criteria based on proposed cut‐offs in CSF or brain imaging measures: participants were classified on the AD continuum (A+T‐N‐, A+T+N‐, A+T‐N+, A+T+N+) and non‐AD continuum (all other subgroups); and 2) using a machine learning (ML) based method (ensemble linear discriminant model built using CN and AD participants) to classify MCI participants as CN‐like or AD‐like based on continuous CSF biomarkers, regional amyloid PET, FDG‐PET, and volumetric MRI (variables contributing to AT[N] classification; Table 1). Predictive performance of AT[N]‐based classification and ML‐based classification were assessed using clinical outcomes at 0.5, 1, 2, 3, 4, and 5 years of follow‐up.ResultA total of 414 aMCI participants had at least one wave of follow‐up (N of 397, 400, 346, 297, 238, and 93 at 0.5, 1, 2, 3, 4, and 5 years, respectively). The progression rate from aMCI to AD was 5.8% at 6m, 9.5% at 1y, 17.9% at 2y, 19.9% at 3y, 20.2% at 4y, and 15% at 5y of follow‐up. Presence of AD pathology at baseline (AT[N]‐based AD continuum) predicted progression to AD with high sensitivity but modest specificity during follow up (Table 2). ML‐based classification was less sensitive but substantially more specific. ML‐based classification models were significantly more accurate than AT[N]‐based classification at all follow‐up periods (McNemar test, p‐value<0.001 for all).ConclusionA single cut‐point approach used in AT[N] classification lacks accuracy when research questions require high diagnostic or prognostic certainty. A multivariate approach using ML methods can improve accuracy of prediction of disease progression. PPVs were higher at all follow‐up time points for ML‐based classification, indicating that progression to AD occurs at a higher rate in persons identified by ML.

AbstractBackgroundRecent evidence has implicated brain networks in the pathophysiology of Alzheimer’s disease (AD). However, the phenotypic variability in AD, inter‐individual differences in tau PET uptake patterns, and dynamic nature of functional connectivity have been relatively neglected. In this talk, I will discuss our recent work linking measures of dynamic brain organization to spatially independent tau PET patterns defined in a phenotypically diverse cohort.MethodsWe identified all Mayo Clinic participants with concurrent amyloid PET (PiB), tau PET (AV1451) and task‐free fMRI (impaired participants were all Aβ+). We used independent component analysis (ICA) to identify spatially independent patterns of tau PET uptake. For the fMRI data, we used a Hidden‐Markov Model (HMM) design to identify distinct patterns of activity/connectivity (‘states’). The ICA and HMM parameters were used to project/segment an independent cohort (ADNI) and subsequent longitudinal timepoints for the Mayo cohort. Multilevel Bayesian regression models were fit in Stan to explore the association between phenotype, component load and brain state.ResultsThe optimal HMM model identified 10 brain states (Figure 1). The optimal ICA decomposition identified 42 tau PET components, of which 13 were potentially related to AD pathobiology (Figure 1). Component loads and brain state dwell time differed between diagnostic groups in both Mayo and ADNI cohorts. One state was much more likely to occur in AD in both cohorts (State 8). Cross‐sectionally, State 8 dwell time was associated with higher tau loads, while higher baseline dwell‐time was predictive of ∆tau in participants with at least 2 timepoints (Figure 2).ConclusionWe found links between dynamics of functional connectivity, tau deposition, and cognitive impairment across the AD spectrum using two different cohorts and longitudinal data. Specifically: 1) our method for characterizing tau PET uptake was able to capture the magnitude and spatial pattern of binding across diverse phenotypes; 2) Our data supports a role for dynamic brain connectivity in cognitive impairment due to Alzheimer’s disease; 3) Dynamic connectivity appears to be directly related to AD pathobiology as measured by tau PET. Incorporating the time‐varying nature of connectivity will be important in future studies of connectome‐based disease progression in AD.

AbstractBackgroundA previous study demonstrated that cerebrospinal fluid (CSF) phosphorylated tau 217 (pTau217), compared to phosphorylated tau 181 (pTau181), correlated more strongly with Tau PET and was better able to distinguish between AD and non‐AD neurodegenerative diseases. The aim of this study was to compare these two CSF tau markers for cognitive progression.MethodWe included 666 cognitively unimpaired (CU) and 89 mild cognitive impairment (MCI) Mayo Clinic Study on Aging community‐based participants (mean age of 71.6 years) with CSF amyloid and tau measurements. CSF amyloid‐beta 42 (Aβ42) was measured with automated electrochemiluminescence Elecsys immunoassays (Roche Diagnostics); a cut‐off of <1026 pg/ml was considered amyloid positive (A+). CSF pTau181 and pTau217 were measured with the Mesoscale discovery platform (MSD). Cutoffs of pTau181≥217 pg/ml and pTau217≥288 pg/ml were considered tau positive (T+). Cox proportional hazards models with time as the time scale were used to examine the relationship between the CSF markers and progression from CU to MCI and MCI to dementia. Covariates included age, sex, education, and APOE.ResultTau positivity (T+) based on pTau217 (HR = 2.45, 95% CI: 1.45, 4.11) was a greater risk factor for incident MCI compared to pTau181 (HR = 2.07, 95% CI: 1.19, 3.61). Among individuals who were A+T+, T+ based on pTau217 (HR = 3.01, 95% CI: 1.66‐5.45) was a greater risk factor of incident MCI than pTau181 (2.80, 95% CI: 1.48, 5.31). Among the 89 MCI participants, T+ based on pTau217 was a much greater risk factor for dementia (HR = 4.92, 95% CI: 1.37, 17.61) than pTau181 (HR = 1.55, 95% CI: 0.43‐5.55).ConclusionCSF pTau217 is a better marker of cognitive progression, especially from MCI to dementia, compared to CSF pTau181.

AbstractBackgroundWhile β‐amyloid and tau PET imaging are widely studied in cognitive aging and Alzheimer’s disease (AD), little is known about findings in aging multiple sclerosis (MS) patients. As aging is an essential factor in developing neurodegeneration in MS, β‐amyloid and tau biomarkers may be utilized to understand the influence of AD pathophysiology on cognitive aging in MS.MethodWe studied patients with MS (n=16) and controls (n=80) matched for age, sex and APOE ε4 status from the population‐based Mayo Clinic Study of Aging who underwent Pittsburgh compound‐B (PiB) PET imaging. Twelve of the 16 patients with MS and 60 matching controls also underwent flortaucipir (AV1451) tau PET. Cortical PiB and AV1451 standard uptake value ratios (SUVr) from previously determined AD signature regions and from the entire cortex were calculated for group comparisons and testing for associations with age.ResultAD signature PiB SUVr (p=0.044), total cortical PiB SUVr (p=0.048) and the frequency of abnormal PiB SUVrs (p=0.040) were lower in patients with MS than controls. Although AD signature and total cortical AV1451 SUVrs were not different between the groups, the frequency of abnormal AV1451 SUVrs was higher (p=0.041) in patients with MS than controls. The association of AD signature PiB SUVr with age was steeper in the controls compared to patients with MS (p=0.002). Similarly, the association of total cortical PiB SUVr with age was steeper in the controls compared to patients with MS (p=0.002). There was no difference in the association of AV1451 SUVr findings with age between the MS patients and controls (Figure 1). The AD signature PiB SUVr increased more with increasing AD signature AV1451 SUVr in controls than in patients with MS (p=0.03) (Figure 2).ConclusionIn patients with MS, cortical β‐amyloid deposition was lower than age‐matched controls, while cortical tau deposition was not different between the groups. This suggests that some aspect of MS pathobiology retards the accumulation of β‐amyloid, but not the accumulation of tau.

AbstractBackgroundAlzheimer’s disease (AD) is marked by selective disruptions to anatomically‐distinct brain regions and their connectivity which collectively support cognition. Based on a variation of pattern‐recognition analysis of FDG‐PET images, we previously developed a framework, the global functional state space (GFSS), which captures this macroscale architecture by assigning each individual a series of weightings on 10 independent patterns (eigenbrains=EBs). In this study, we applied the GFSS as a quantitative endophenotype to test whether known AD genetic risk factors were associated with variability in large‐scale neurodynamics represented by the GFSS.MethodThe study sample included 410 participants from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) with evidence of cognitive impairment (CDR‐global>0) and amyloidosis (positive amyloid‐PET). The GFSS was estimated for each individual using the set of 10 EBs to describe global patterns of variation in glucose metabolism. AD genetic risk variants identified in prior large consortium case‐control studies were included for analysis if the variant (or a proxy by linkage disequilibrium r 2>0.7) was assayed on both genotyping platforms used in the sample and if the minor allele frequency exceeded 1%. Genetic associations with the collective set of EBs were tested using multivariate linear regression under a dominant genetic model.ResultSignificant multivariate associations were identified for CD33 (p=7.04x10‐4), MEF2C (p=0.001), and SORL1 (p=0.02). These findings remained significant after covarying for age and sex. The strongest component loadings for CD33 (p=9.09x10‐4; std. β=‐0.16) and MEF2C (p=1.59x10‐4; std. β=‐0.18) were for EB1, with the minor allele in each case associated with a negative axis deviation in the direction of the GFSS area representing episodic memory dysfunction. SORL1 displayed its strongest loading onto EB4 (p=0.03, std. β=‐0.11). The association of APOE ɛ4 with the GFSS was marginally nonsignificant (p=0.09).ConclusionWe identified associations of risk variants in CD33, MEF2C, and SORL1 with differences in patterns of FDG hypometabolism in persons with symptomatic AD. These findings establish the GFSS as a useful endophenotype for genetic analyses, particularly as an avenue toward a mechanistic understanding of how heritable factors may influence inter‐individual variability and syndromic diversity in AD. Further investigation in larger and independent cohorts is warranted.

AbstractBackgroundNeurodegeneration of the substantia nigra in dementia with Lewy bodies is associated with iron deposition, which increases the magnetic susceptibility of the substantia nigra on MRI. The objective of this study was to measure the magnetic susceptibility of the substantia nigra in patients with dementia with Lewy bodies (DLB) and patients with mild cognitive impairment with at least one core feature of DLB (MCI‐DLB) compared to cognitively unimpaired controls (CU) by quantitative susceptibility mapping (QSM).MethodWe studied clinically diagnosed patients with DLB (n=36), patients with MCI‐DLB (n=16), and age‐, gender‐matched CU (n=104) in a case‐control study. MR images used for QSM were performed on 3T MRI with a multi‐echo (5 echoes) 3D gradient recalled echo sequence with a bipolar‐readout. Three‐dimensional regions of interest (ROIs) of bilateral substantia nigra, based on the DISTAL atlas, were propagated from the Mayo Clinic Adult Lifespan Template (MCALT) to each subject’s T1 image using ANTs nonlinear registration. The mean susceptibility values of the substantia nigra were extracted by aligning each subjects QSM image to their T1 image and corresponding ROIs with a 6 DOF rigid registration.ResultThe mean susceptibility values of the substantia nigra bilaterally was 0.110 ± 0.026 for DLB, 0.099 ± 0.028 for MCI‐DLB and 0.091 ± 0.025 for controls (Figure 1). No difference was found across the mean susceptibility from left, right and combined bilateral substantia nigra. Compared to controls, patients with DLB had higher susceptibility in the substantia nigra (p<0.001). The susceptibility of substantia nigra showed an increasing trend from controls to MCI‐DLB and to DLB (p<0.001). There was no correlation between the Unified Parkinson's Disease Rating Scale (UPDRS) and the mean susceptibility in the substantia nigra in patients with MCI‐DLB and DLB (r=0.03; p=0.82).ConclusionQuantitative susceptibility mapping is sensitive to the increased magnetic susceptibility due to higher iron content in the substantia nigra in DLB. The trend of increasing susceptibility from CU to MCI‐DLB and to DLB suggests that iron deposition in the substantia nigra starts to increase as early as the prodromal stage in DLB and continues to increase as the disease progresses independent of the severity of Parkinsonism.

AbstractBackgroundWhite matter (WM) hypointensity volume in T1 MRI correlates strongly with T2 hyperintensities and both are typically associated with advancing age and vascular brain injury (VBI). Recent research has suggested that Alzheimer’s Disease (AD) neuropathologic changes (ADNC) also increases WM hypointensity volume. Sporadic early onset AD (EOAD) patients provide a unique model to examine the relationship between ADNC biomarkers and WM hypointensity volume in the absence of advanced age, cardiovascular injury, or autosomal dominant genetic mutations.MethodIncluded in the analysis were sporadic EOAD patients (n = 75) and cognitively normal controls (n = 39) from the Longitudinal Alzheimer’s Disease Study (LEADS) for whom full demographic and health information were available. WM hypointensity volume in T1 MRI was segmented by Freesurfer and corrected for total white matter volume. Multiple‐factor ANOVA models with logarithmic transformations were used to determine the effects of age, cohort, and presence of cardiovascular conditions on WM hypointensity volume. Further multiple‐factor ANOVA models assessed whether the addition of covariates improved the model fit.ResultIn a reproduction of prior research, WM hypointensity volume was correlated with age for controls (p < 0.05), but not for EOAD patients. We additionally found an effect of cohort, such that EOAD patients had a higher mean WM hypointensity volume than controls when accounting for age (p < 0.05). No difference in the presence or number of cardiovascular conditions was found between groups. Additionally, no correlation was found between WM hypointensity volume and the presence or number of cardiovascular conditions. Including cardiovascular condition as a covariate failed to improve the model fit, even when accounting for interaction effects.ConclusionThese results support the hypothesis that patients with ADNC have higher WM hypointensity volume beyond what would be expected based on aging or cardiovascular risk factors. Further research is needed to assess the co‐localization of WM hypointensities with cortical degradation, amyloid pathology, and tau pathology.

AbstractBackgroundData shared for research use must be de‐identified to protect participant confidentiality. A recent study has shown that face imagery from brain imaging studies could be used to re‐identify anonymous participants via automatic face recognition, linking them with all study‐collected protected health information. Here we measure how well existing software for removing MRI face imagery protects against this privacy threat.MethodWe recruited 153 participants with existing 3D FLAIR MRI and photographed their face from 5 varying angles. We used surf_ice to reconstruct each participant’s face from their MRI and tested whether free, publicly‐available face recognition software (Microsoft Azure) could match their photos to their correct MRI. We tested two existing programs for removing face voxels in MRI: mri_deface (FreeSurfer) and pydeface (python package), and one in‐house implementation. We observed that existing methods often left participants’ eyes intact while removing most of their nose/mouth/chin. This prevented Azure from detecting the reconstructions as faces, therefore preventing its attempt of face recognition. We hypothesized that the retained eye imagery alone could be sufficient for face recognition, if a face were detected. Therefore, we also reconstructed each de‐faced MRI after applying novel pre‐processing that replaced any removed/missing face voxels with those from our population‐average template, e.g. replacing missing noses/mouths/chins with population‐average noses/mouths/chins, before testing face recognition.ResultsWithout any MRI de‐facing, Microsoft Azure correctly matched 78% of participants’ photos to their MRI. After mri_deface and pydeface, 8% and 6% were still identified, respectively. After preprocessing the de‐faced MRI to replace missing areas with those from a population‐average template prior to face reconstruction, the face imagery remaining in the de‐faced images was enough to allow re‐identification of 23% and 20% of participants, respectively. With an in‐house face removal implementation that removes the whole face from every image, 0% of faces were re‐identified from the standard reconstructions, and 5% were identified from the “missing‐parts‐replacement” reconstructions.ConclusionsEven after applying popular MRI de‐facing software, Microsoft Azure face recognition automatically re‐identified up to 23% of participants. Better de‐facing software is needed to remove more of the face more reliably, without affecting brain measurements from de‐faced images.