Vemuri, Prashanthi

AbstractBackgroundResearch subjects can potentially be re‐identified from de‐identified MRI, CT, and PET brain scans with up to 98% accuracy using Microsoft Azure’s cloud‐based commercial facial recognition software. This showed the need to “de‐face” publicly shared research brain scans. Subsequently, Microsoft has begun restricting its face recognition services, intending to prevent misuse. This study tests a variety of popular open‐source computer vision and facial recognition software packages to measure the re‐identification risk with only unrestricted free software.MethodThis study used brain MRI from Siemens 3T scanners and face photographs from 182 participants (ages 30‐90+) in the Mayo Clinic Study of Aging. Each participant had photos taken with an iPad, which were cropped loosely around the head and converted to black and white. A total of 81 renders (2D reconstruction images; Figure 1) were made from each 3D T2‐FLAIR MRI, simulating various camera and lighting positions. These face renders approximate black and white photos, allowing use with software packages developed for matching faces/features between photographs.Open‐source Python packages (OpenCV, DeepFace, face_recognition, scikit‐learn) and their included pre‐trained models (designed for photographs) were used to match participant photographs to the labelled library of MRI‐based face renders. A correct match was recorded when the top‐scoring render or set of renders automatically chosen by the algorithm was from the MRI of the correct participant.ResultFeature matching with ORB and SWIFT (both from OpenCV) produced no correct matches, despite a prior publication reporting moderate success. The best match rates (Figure 2) were 58‐59% accuracy using face_recognition or DeepFace packages with support vector machines (SVMs). Without SVMs, these packages performed modestly worse (51%‐55%).ConclusionWhile the highest match rates obtained by widely available open‐source packages (roughly 60%) are much lower than those of a limited‐access commercial product (Microsoft Azure, with 98%), they show that actors with only access to no‐cost readily available software could still re‐identify a research subject from a brain scan most of the time. This further highlights the need for image refacing software, such as mri_reface, to protect the privacy and identifies of research volunteers.

AbstractBackgroundPlasma Aβ42/40 and p‐tau217 can predict amyloid positivity in cross‐sectional studies. However, it is unclear how plasma biomarkers perform longitudinally, which is important to inform their utility in tracking disease progression. The goal of this study is to describe temporal evolutions of plasma Aβ42/40 and ptau217 ratio (p‐tau217r) measured via mass spectrometry, p‐tau217 measured via an immunoassay, and amyloid PET.MethodsWe included participants in the Mayo Clinic Study of Aging with amyloid PET (PiB) and plasma Aβ42/40 + ptau217r (mass spectrometry, C2N) and/or p‐tau217 (Lilly, MSD). We fit a joint non‐linear mixed effects model with Aβ42/40, p‐tau217r, p‐tau217 (Lilly), and amyloid PET as outcomes and age, sex, APOE ε4 carriership and education as covariates. The model included a per‐participant adjustment (time‐shift) for each biomarker indicating how much earlier or later in time the biomarker was estimated to progress relative to the average participant, a correlation (R) between adjustments for each pair of biomarkers, and the impact of covariates on each biomarker’s timing.ResultsThe study included 1459 participants, age 77 (10) years, 44% female, 82% cognitively unimpaired. Mean (SD) years of followup was 5.6(3.0) years for mass spectrometry plasma measures, 2.5(0.7) years for p‐tau217 (Lilly), and 4.6(3.3) for amyloid PET. Model fits are shown for each biomarker, with greater variability in serial measurements for plasma than PET (Figure 1). Individual adjustments between each of the plasma biomarkers and amyloid PET were highly correlated: p‐tau217r, R=0.73 (95%CI: 0.67, 0.78); p‐tau217 (Lilly), R=0.66 (0.62, 0.7); Aβ42/40, R=0.6 (0.52, 0.68) (Figure 2). The p‐tau217r and Aβ42/40 individual adjustments were modestly correlated, R=0.39 (0.28, 0.49). APOE ε4 carriership had the strongest covariate effect on all biomarkers, e.g. p‐tau217r progression 9.4 (6.7, 12) years earlier (Figure 3).ConclusionsPlasma Aβ42/40 and p‐tau217 measured with currently available assays may both be amenable to longitudinal modeling, although the degree of variability in Aβ42/40 over time may pose challenges. Findings suggest that both plasma Aβ42/40 and p‐tau217r may provide information regarding an individual’s timing of progression on amyloid PET; lower correlations with each other than amyloid PET suggests that each analyte may provide complimentary information.

AbstractBackgroundVascular contributions to cognitive impairment and dementia (VCID) are often comorbid with Alzheimer’s disease and increase the risk of dementia. Blood‐based biomarkers may be promising for identifying individuals at high risk for VCID due to small vessel disease (SVD).MethodWe included 1709 participants from the Mayo Clinic Study of Aging who had concurrent MRI, plasma neurofilament light chain (NfL) and glial fibrillary acidic protein (GFAP) measured on the HD‐X Simoa Quanterix platform, and clinical dementia rating scale (CDR). We categorized participants into SVD vs. not using WMH/TIV≥1.3% which corresponds to Fazekas>=2 cutoff and cognitively impaired (CI) vs. not using CDR≥0.5. We obtained four groups using SVD and cognitively normal and impaired classifications: cognitively unimpaired without SVD (CU), CU with SVD (Vas‐CU), CI with SVD (VCI), and CI without SVD (Non‐vas‐CI). Univariate and multivariable logistic regression models were used to obtain odds ratios, confidence intervals and ROC curves to evaluate the association between plasma markers and each disease group compared to CU. We also evaluated these analyses in amyloid positive and negative participants.ResultThere was a high proportion of CU (Vas‐CU/(CU+Vas‐CU)=18%) and CI (VCI/(VCI+ Non‐vas‐CI)=47%) participants with SVD in the population‐based sample. Higher plasma levels were associated with greater vascular burden and differed by amyloid status (Figure 1). ROC analyses showed that both NfL and GFAP provided discrimination between Vas‐CU and CU groups (AUC=0.83) as well as VCI and CU groups (AUC=0.88) (Figure 2, Table 1). Though GFAP and NfL are non‐specific and may also reflect non‐vascular damage, we found that these plasma markers did not clearly separate NonVas‐CI from the CU group in the overall cohort (AUC=0.69). When amyloid status was available, the discriminative power of NfL and GFAP in separating the 4 groups was very low in amyloid positive participants. On the other hand, NfL and GFAP separated Vas‐CU and VCI groups from CI reasonably well in the amyloid negative group.ConclusionPlasma biomarkers associated with SVD‐burden provided reliable discriminative power in predicting VCID, particularly in the absence of amyloid, suggesting their utility in the prognosis and stratification for clinical trials.

AbstractBackgroundQuantifying white matter using diffusion MRI (dMRI) has been proposed for measuring early microstructural tissue changes due to cerebral small vessel disease and aid in quantifying vascular contributions to cognitive impairment and dementia (VCID). Our goal was to compare the usefulness of longitudinal white matter changes in the commonly available diffusion MRI measures for VCID prevention trials.MethodWe included 718 participants over 50 years of age (mean age: 71.1(9.6) years) from the Mayo Clinic Study of Aging, a population‐based sample, with at least two dMRI scans and structural imaging. We computed the commonly available single‐shell dMRI measures (fractional anisotropy, mean diffusivity, and two MarkVCID measures ‐ free water and peak‐width skeletonized mean diffusivity [PSMD]) at each time point. We tested for voxel‐wise associations between dMRI markers and vascular risk measured by cardiovascular metabolic condition (CMC) and observed a region‐specific dependance across all dMRI measures. Using both global and regional dMRI measures, we (i) examined the longitudinal association of dMRI measures with cognition and (ii) computed sample size estimates for a hypothetical clinical trial. We also included white matter hyperintensities (WMH) and our previously proposed composite vascular white matter score (combination of WMH and fractional anisotropy of the genu) as a comparison.ResultVascular risk was associated with all single shell dMRI measures in the genu of the corpus callosum, which we included as a regional dMRI marker for comparison (Figure 1). All dMRI markers correlated with cognitive performance longitudinally (Table 1) and had comparable sample size estimates required for hypothetical VCID clinical trials (Figure 2). Further, global free water and the composite vascular white matter score had the smallest sample size estimates.ConclusionAll commonly used dMRI markers had significant frontal lobe changes due to vascular risk. Both global and regional corpus callosum dMRI markers were sensitive to longitudinal cognitive decline, suggesting their utility in measuring the slowing down of VCID. The composite vascular white matter score, global free water, and WMH show promise as VCID biomarkers. Further work is needed to validate these markers on multiple populations.

AbstractBackgroundThere has been a recent proliferation of various quantities based on single shell Diffusion Tensor Imaging (DTI) for capturing overall cerebrovascular health, especially Small Vessel Disease (SVD) as a single number. The existing literature has gaps in the comparison of these measures, so we evaluated them as predictors of both current and future White Matter Hyperintensity (WMH) as a proxy of SVD.MethodUsing baseline 3T MRI examinations (T1, DTI, FLAIR scans) of 598 participants (>60 years, amyloid negative, to limit to those on the SVD pathway) from the Mayo Clinic Study of Aging, we calculated five summary DTI quantities with varying degree of complexity. We estimated their partial R2 from cross‐sectional linear regression models and computed semi‐partial R2 from mixed models predicting WMH volume as a percentage of Total Intracranial Volume (TIV). All models were adjusted for age, sex, and scanner model, and the WMH and based were log‐transformed before fitting. We also conducted subset analyses with high (WMH+) vs. low (WMH‐) baseline WMH load (cut at 1.62%) to compare the performance of the measures in early vs. widespread WMH.ResultAll MD‐based quantities computed on the whole brain, including FWF, performed similarly to each other, but Genu FA underperformed at explaining WMH variance (Table 1). PSMD (skeletonized or atlas based) was more sensitive in WMH+ but dramatically less so in WMH‐. In WMH‐ and overall cohort, FWF and JHU MD (which is a simple averaging of MD in the JHU atlas) performed well.ConclusionAll MD‐based measures performed similarly to each other in the whole dataset but dichotomizing by WMH load showed interesting differences. PSMD measurements are driven by the extreme end (95th percentile) of the MD distribution and therefore more sensitive in WMH+. Genu FA, a regional measure, was not useful for predicting whole brain WMH change. We noticed some variation in how the quantities are affected by scanner model (GE Signa HDxt and Discovery 750) which needs broader investigation. Simply averaging MD over WM avoids additional processing and is more sensitive to WM degeneration that is not (yet) severe.

AbstractBackgroundDementia with Lewy bodies (DLB) frequently co‐occurs with Alzheimer’s disease (AD) pathologies, exacerbating disease progression. Biophysical models of diffusion imaging data, such as Neurite Orientation Dispersion and Density Imaging (NODDI), may reveal novel insights into the neurobiological substrates of AD on cortical and white matter microstructural injury.MethodA cohort of 57 DLB patients on the DLB spectrum (33 clinically probable DLB and 14 prodromal DLB) and 57 cognitively unimpaired (CU) controls underwent NODDI and PET imaging with [18F]‐Flortaucipir and [11C]‐PiB (Table 1). NODDI parameters were derived from white matter tracts and grey matter regions using the JHU “Eve” and Mayo Clinic Adult Lifespan Template atlases, respectively. Conditional logistic models were used to compare these parameters between the two groups, accounting for the 1:1 matching on age and sex. The associations of tau and amyloid‐β with NODDI parameters across the white matter tracts were examined with linear regressions. A region‐by‐region approach was adopted to evaluate the relationships of tau and amyloid‐β with grey matter NODDI, adjusting for age, APOE, and local grey matter volumes.ResultDLB spectrum patients exhibited extensive white matter abnormalities relative to CU, including lower fractional anisotropy, higher mean diffusivity (MD), and lower tissue‐weighted neurite density (tNDI). In grey matter, these deficits were paralleled with higher MD and lower tNDI and ODI, primarily in the temporoparietal and posterior cortices. Regression models revealed significant associations of elevated [18F]‐Flortaucipir uptake with deficits in temporal and limbic white matter tracts and a temporal‐lobe pattern of higher MD and lower tNDI in the cortex (Figure 1). No associations were observed with [11C]‐PiB in the DLB spectrum.ConclusionThese findings demonstrated an in vivo relationship between tau deposition and neurite deficits that spatially aligned with the canonical AD “cortical signature” of neurodegeneration in both white matter tracts and grey matter regions in DLB. Our study underscores the detrimental impact of AD pathology in DLB and provides a rationale for further investigations into the underlying mechanisms of AD in DLB. Future longitudinal studies tracking the progression of these changes could inform therapeutic strategies for DLB in the presence of AD pathologies

AbstractBackgroundLarge‐scale studies comparing sporadic early‐onset AD (EOAD, age<65) and late‐onset AD (LOAD, age≥65) are lacking. We compared amyloid‐PET outcomes (positivity rate and amyloid burden) between patients clinically diagnosed with sporadic EOAD vs LOAD, leveraging data from the Longitudinal Early‐Onset AD Study (LEADS) and the Alzheimer’s Disease Neuroimaging Initiative 3 (ADNI3).Method731 patients meeting the 2011 NIA‐AA criteria for AD dementia or MCI were included (505 early‐onset from LEADS, 226 late‐onset from ADNI3, Table 1). All participants underwent amyloid‐PET with [18F]Florbetaben or [18F]Florbetapir. Amyloid positivity was centrally determined by a process involving a visual read by a trained expert and PET‐only quantification; in case of a discrepancy, a read from an independent physician acted as a tiebreaker. Logistic regressions in each cohort examined relations between amyloid positivity and age, sex, MMSE and APOE4 genotype. Amyloid burden was independently quantified in Centiloids using an MRI‐based pipeline. Mean Centiloids in LEADS and ADNI were compared with two‐way ANOVA, for visually positive and visually negative scans.ResultAmyloid positivity rate was higher in LEADS (76%) than ADNI (64%, p<0.001, Figure 1A). Lower MMSE and APOE4 genotype increased odds of amyloid positivity in both cohorts, although the APOE4 effect was stronger in ADNI than LEADS (OR=10.1 versus 2.4, p=0.007, Table 2). Amyloid positivity was more common in females across cohorts, but this effect was only statistically significant in LEADS (Table 2). Centiloids were bimodally distributed in both cohorts, although the separation between positive and negative scans was more prominent in LEADS (Figure 1B). Visually positive scans had significantly higher Centiloids in LEADS than in ADNI, whereas no cohort difference was observed for visually negative scans (Figure 1C). Sensitivity analyses showed that this effect was driven by patients with MCI (CDR≤0.5; Figure 1D‐E).ConclusionThe lower amyloid positivity rate in ADNI might be due to AD‐mimicking pathologies being more common at an older age. The higher amyloid burden in early‐onset, amyloid‐positive patients could reflect younger patients being diagnosed later in the disease course compared to typical, late‐onset patients. Alternatively, younger patients might tolerate higher neuropathology burden due to higher brain reserve or fewer co‐pathologies.

AbstractBackgroundUnderstanding the relationship between AβPET and AβCSF biomarkers will define their potential utility in Aβ treatment. Few longitudinal or neuropathological comparisons have been reported. We assessed the relationship of AβPET and AβCSF biomarkers in a large community cohort.MethodParticipants 50+ years with coincident AβPET and AβCSF biomarkers (pTau181/Aβ42, n=505, and Aβ42/40, n=54) were included. Of these participants, a subgroup was identified that had neuropathology data (n=47) of which many had longitudinal AβPET and CSF testing. The relationships of AβPET and CSF biomarkers were assessed cross‐sectionally in all participants and longitudinally in those with autopsy data. Analyses were performed using AβPET Meta‐ROI and AβPET early‐Aβ‐deposition ROIs (ROIs with early Aβ deposition: fusiform, angular, inferior temporal, middle temporal, middle occipital, and calcarine).ResultIn biomarker tests performed at the same time, more participants were AβPET+ and CSF‐ than AβPET‐ and CSF+ using Meta‐ROI AβPET. This discordance occurred more frequently when using early‐Aβ‐deposition regions (Figure 1). Of 47 participants with autopsy data, 35 had Thal stage ≥1. The sensitivity for earliest detection of cortical Aβ neuropathology by AβPET was 89% (31/35) and by CSF was 66%, (23/35). Eight of these 35 had positive AβPET findings and negative coincident CSF tests at the first time tested. Of these, 5/8 were Meta‐ROI and early AβPET ROI positive. The other 3/8 were only positive on early‐Aβ AβPET ROIs. Four of the 8 with negative CSF and positive AβPET had subsequent testing and 2/4 became positive on CSF tests after 1.5 years while 2/4 continued to be negative on subsequent testing.ConclusionAβPET is more sensitive for detection of early cortical Aβ deposition than CSF biomarkers in cross‐sectional or autopsy data. While the assessment of AβPET by standard Meta‐ROI AβPET methods showed cortical Aβ deposition occurring when not indicated by CSF biomarkers, AβPET ROIs for early Aβ deposition detected additional participants that were AβPET positive and AβCSF negative. These findings have implications for detecting early Aβ neuropathology among CU participants in Aβ prevention trials when robust detection is needed. Further work is needed to compare Aβ plasma biomarkers and AβPET using these methods.

AbstractBackgroundSeveral programs are available for automatically “de‐facing” brain images. These are designed to prevent the potential use of face imagery to re‐identify research participants. Previous works have compared available software and mri_reface is frequently among the top methods. The “Minimally Invasive DeFacing of MRI images” (MiDeFace) program was more recently added to FreeSurfer, but minimal validation of MiDeFace has been published.MethodWe ran mri_reface 0.3.3 and MiDeFace 7.3.2 on two datasets: A) 207 participants with face photographs and 3D T1‐weighted and 3D T2‐weighted‐FLAIR MRI from the Mayo Clinic Study of Aging (all unimpaired, Siemens); and B) 300 participants from ADNI3 with 3D T1‐weighted MRI (50 unimpaired + 50 with clinical AD, from each of 3 vendors). For dataset A, we tested the rate at which Microsoft Azure face recognition matched each participant’s photographs with the correct MRI, comparing unmodified images, those from mri_reface, and those from MiDeFace. For dataset B, we compared measured regional brain volume and cortical thickness from original vs. de‐faced images with SPM12, FreeSurfer 7.4.1, and the FSL‐UK Biobank pipeline. For both datasets, images were visually rated to assess removal of the face and retention of brain tissue.ResultAutomated face recognition match rates were lower with mri_reface (17/207 8.2%) than with MiDeFace (24/207 11.6%), indicating stronger identity protection with mri_reface, but both were greatly reduced vs. unmodified images (203/207 98.1%). Example images with both programs are shown in Figure 1. From visual ratings (Table 1), rates of complete success (modification of the entire face while retaining the entire brain) were much higher with mri_reface, but the great majority of failures from MiDeFace were subtle and occurred in relatively less recognizable face regions. The effects of both programs on downstream imaging measurements (Figure 2) were small and very comparable, but MiDeFace was favored mildly for SPM12 and for FSL, while mri_reface was more favored for FreeSurfer.ConclusionAltogether, mri_reface more reliably removed the complete face, leading to smaller rates of potential re‐identification (8.2% vs 11.6%). Effects on downstream image measurements were small and comparable with both methods.

Abstract Background The Mayo Clinic Study of Aging (MCSA) is a longitudinal, population‐based study of residents of Olmsted County, Minnesota, USA. MCSA is releasing de‐identified clinical and imaging data on GAAIN.org to benefit the research community. Method We included longitudinal clinical data from all participants of the Mayo Clinic Study of Aging 30‐90 years of age (average 70.0y) through the first 15 years of the study. Clinical data includes age, sex, self‐reported race/ethnicity, APoE4 allele status, cognitive impairment status (normal, MCI, dementia, or other), height, weight, blood pressure, multi‐domain neuropsychological scores, neuroimaging summary measurements, medications, cardiovascular risk factors, social activity, sleep, mood/anxiety, functional measures, and more. For our initial release of imaging data, we selected the first imaging timepoint acquired on 3T GE MRI scanners and amyloid (PiB) PET during the same visit, from all available imaging participants 50‐90 years of age (average 70.7y). Currently, we included 3D T1‐weighted (MPRAGE) MRI (1x1x1.2mm 3 ), T2‐weighted FLAIR MRI (0.85x0.85x3.6mm 3 ), diffusion MRI (2.7mm resolution with 41 directions at b=1000 and 5 b=0), and numeric cortical global SUVR and Centiloid values from PiB PET scans at 40‐60 minutes post‐injection. All images are released in DICOM format and thoroughly de‐identified using CTP while ensuring preservation of necessary fields for analyses. T1‐weighted and T2‐FLAIR MRI were also de‐faced (identifiable face imagery was removed and replaced with an average) using mri_reface version 0.3.3 to further protect participants’ privacy, and all de‐faced images were visually inspected by a trained data scientist for quality assurance. Result The first phase of our public data release includes clinical data from 5925 unique participants from the Mayo Clinic Study of Aging (MCSA), each ranging from 1‐12 visits at roughly 15‐month intervals, and three MRI sequences (T1‐w, T2‐w‐FLAIR, and diffusion) from 1802 unique participants. Demographic summaries are given in Table 1. Conclusion This large, richly characterized, de‐identified clinical and imaging dataset is available to the research community at https://www.gaaindata.org/partner/MCSA , and it is suitable for a wide range of research uses. Additional data will be released over time, including de‐faced amyloid PET images.