Citations

AbstractBackgroundMany proposed clinical decision support systems (CDSS) require multiple disparate data elements as input, which makes implementation difficult, and furthermore have a black‐box nature leading to low interpretability. Fluorodeoxyglucose Positron Emission Tomography (FDG‐PET) is an established modality for the diagnosis of dementia, and a CDSS that uses only an FDG‐PET image to produce a reliable and understandable result would ease both of these challenges to clinical application.MethodA deep variational autoencoder (VAE) was used to extract a latent representation of each image through prior training from FDG‐PET brain images (n=2000). This unsupervised VAE has a novel graph convolutional architecture that makes it applicable to masked template space images. A logistic regression model was used to classify each image. A parametric study of the latent space revealed the imaging features that were used in the logistic regression model to differentiate each class. A separate dataset of participants labeled with their clinical diagnosis (n=1239) was used to assess the logistic regression model’s diagnostic accuracy of cognitively unimpaired (CU) (n=679) and between common dementia subtypes: Alzheimer’s disease (AD) (n=310), Lewy body dementia (DLB) (n=151), and behavioral variant frontotemporal dementia (bvFTD) (n=99). The logistic regression classifier was evaluated using receiver‐operator characteristic (ROC‐AUC) curves.ResultROC‐AUC curves for CU and each dementia subtype are illustrated in Figure 1a. A k nearest neighbors model was used to develop a graphical representation of the latent space where nodes are images and edges are drawn between nearest neighbors, illustrated in Figure 1b. The graph of the latent space demonstrates a comparison between dementia subtypes and CU images.ConclusionIn this study, we developed a proof‐of‐concept VAE that can achieve differential diagnosis with 84.8% balanced accuracy. The model architecture provides visual interpretability, and shows that the hippocampi, parietal lobes, and cerebellum, were useful for distinguishing AD, the frontal lobe, posterior cingulate cortex, and hippocampi were useful for distinguishing DLB, and prefrontal pole, subcortical structures, and occipital lobe were useful for distinguishing bvFTD. These regions of metabolism align with clinical pathology and metabolic patterns.

AbstractBackgroundRecent advances in automatic face recognition have increased the risk that de‐identified research imaging data could be re‐identified from face imagery in brain scans.MethodAn ADNI committee of independent imaging experts evaluated 11 published techniques for face‐deidentification (“de‐facing”) and selected four algorithms (FSL‐UK Biobank, HCP/XNAT, mri_reface, and BIC) for formal testing using 183 longitudinal scans of 61 racially and ethnically diverse ADNI participants, evaluated by their facial feature removal on 3D rendered surfaces (confirming sufficient privacy protection) and by comparing measurements from ADNI routine image analyses on unmodified vs. de‐faced images (confirming negligible side effects on analyses).ResultEffects of the de‐facing methods on FreeSurfer at UCSF (Figure 1‐top), ASHS at University of Pennsylvania (Figure 1 bottom‐left), and longitudinal atrophy at Mayo Clinic (Figure 1 bottom‐right) were comparable with each other and sufficiently small vs. unmodified scans to recommend de‐facing in ADNI‐4. The FSL and mri_reface methods were more thorough in removing facial features, partly because these also removed ears (Figure 2‐top). mri_reface was recommended for ADNI‐4 primarily because fewer (zero) images had modified brain (Figure 2‐bottom), and ADNI leadership approved de‐facing with mri_reface. ADNI performed a second validation of its latest version on another diverse sample of 100 ADNI‐3 participants. These analyses included: boundary shift interval at University College London, where average differences in measurements were all <0.05%; cortical white matter hyperintensities at UC Davis, where they found “minimum impact” (R>0.97); and others.ConclusionADNI‐4 de‐faces all indicated image types (Table 1) before subsequent pre‐processing, analyses, and public release. Sites upload images to the central repository (LONI), where they are queued to the MRI Core at Mayo Clinic for de‐facing. Image types requiring de‐facing are de‐faced using mri_reface. Trained analysts inspect de‐faced images to confirm complete removal of the face and complete non‐modification of the brain, and any flagged images are re‐processed manually until issues are resolved. De‐faced images are converted back to DICOM and re‐uploaded for further distribution. Image types not requiring de‐facing are indicated to proceed directly to subsequent steps. In summary, ADNI‐4 has adopted mri_reface because it successfully de‐faces without substantially affecting brain measurements.

AbstractBackgroundPhase four of the Alzheimer’s Disease Neuroimaging Initiative (ADNI4) began in 2023. This time‐period corresponded to MRI vendors introducing product sequences with compressed sensing (CS), cross‐vendor adoption of arterial spin‐labelling (ASL) and multi‐band slice excitation, and hardware improvements (head‐coils, increased gradient amplitudes). These advances enabled the acquisition of new imaging measures and reduced scan times. The ADNI4 MRI protocol aims to maintain longitudinal consistency across two decades of data acquisition, while adopting new technologies. Here we describe the design and justification behind the ADNI4 protocol and list the target imaging biomarkers.MethodsThe ADNI4 MRI protocol includes 9 MRI sequences; i) T1‐weighted (T1w) MP‐RAGE, ii) a compressed sensing T1w MP‐RAGE, iii) 3D T2w FLAIR, iv) structural isotropic 3D T2w v) T2*‐weighted susceptibility imaging vi) multi‐post labelling delay (PLD) 3D ASL, vii) 2D multi‐band‐multiple‐shell diffusion (dMRI), viii) 2D multi‐band EPI BOLD ix) high resolution, limited field of view, imaging centered on the hippocampus. Table 1 gives a detailed summary of the key protocol parameters and lists some of the imaging biomarkers that can be obtained from each series.ResultsStructural imaging was kept consistent across imaging sites. Variations in vendor specific sequences, software upgrades, and hardware performance capabilities, resulted in variations in advanced imaging strategies. Table 2 highlights the advanced sequences introduced in ADNI4, which will gradually be implemented as the required upgrades and licenses are obtained at each site. The main advanced sequences that are new in ADNI4 from ADNI 3 are i) CS for T1w images, ii) pseudo‐continuous‐ASL (pCASL) being implemented on all 3 vendors (GE, Siemens, Philips), iii) multi‐PLD ASL, which allows for the generation of quantitative maps of cerebral blood‐flow (CBF) and arterial transit times (ATTs), iv) and isotropic 1mm3 resolution for 3D FLAIR. When ADNI4 began, product sequences that utilized artificial intelligence were not widely available and could not be adequately tested before the rollout of ADNI4.ConclusionsThe ADNI4 database will continue to help the neuroimaging community, not only extract valuable imaging biomarkers, but provide the data necessary to test the impact of advanced imaging strategies on diagnostic accuracy and disease sensitivity.

AbstractBackgroundNeurofilament Light Chain (NfL) is a blood biomarker of axonal injury and neurodegeneration that can be used in a variety of neurological disorders. Despite the potential clinical use of plasma NfL across multiple neurological disorders, there is increasing evidence that underlying comorbidities such as renal impairment associated with chronic kidney disease (CKD) and cardiovascular diseases can increase NfL concentrations. The objective of this study was to determine the relationship between plasma NfL concentrations and renal function (CKD staging) in individuals without known neurological conditions.MethodsIndividuals (n=62) seen in an outpatient setting at Mayo Clinic who had various stages of renal impairment (stages 1‐5 by CKD‐EPI 2021 estimated glomerular filtration rate (eGFR) equation) and did not have co‐existing conditions that are known to elevate plasma NfL were retrospectively identified through electronic medical record review. EDTA plasma from these patients collected within +/‐ 30 days of the eGFR determination was identified and stored frozen (‐80°C) until the time of testing. Samples were tested using the Simoa® NF‐light™ V2 Advantage Kit (Quanterix). NfL concentrations in pg/mL were used to calculate age‐adjusted Z‐scores (https://shiny.dkfbasel.ch/baselnflreference/) using an online calculator. Correlation analysis was performed to assess associations of continuous eGFR (mL/min/ 1.73 m2) with plasma NfL age‐adjusted Z‐scores.ResultsA strong negative correlation between NfL Z‐score and eGFR was observed (Spearman’s ρ= ‐0.806). Mean NfL Z‐scores (ranges) in each CKD stage were ‐0.4(‐2.3‐1.4) for stage 1 (N=9), 0.9(‐0.8‐1.9) for stage 2 (N=10), 1.9(‐1.2‐3.4) for stage 3 (N=29), 2.9(2.7‐3.5) for stage 4 (N=6), and 3.2(1.6‐3.7) for stage 5 (N=8); with all groups showing significant differences (P>0.05) except between groups 4 and 5 (P=0.09).ConclusionsPlasma NfL concentrations were increased in individuals with a decreased eGFR. Kidney impairment is an important potential confounding factor to consider when utilizing NfL as a biomarker of neurodegeneration in the clinical setting.

AbstractBackgroundAutosomal dominant Alzheimer disease (ADAD) is characterized by genetic mutations affecting the beta‐amyloid (Aβ) pathway. However, vascular and immune factors play important roles which are not completely understood. Understanding the function of the neurovascular unit (NVU) comprised of neurons, glial cells, and vasculature, at different disease stages appears ideal to developing and evaluating therapeutics. Omics approaches can inform NVU changes and their disease associations.MethodsCSF proteomic data using the Somalogic® platform was generated from 485 participants in the DIAN study, who also had PiB‐PET and MRI assessing Aβ burden, white matter hyperintensity (WMH), and cerebral microhemorrhage (CMH). From previously published AD studies, we identified 33 NVU‐associated proteins with corresponding Somalogic aptamers (Figure 1). We used Principal Component Analysis (PCA) on participants with complete data (n=262, 166 mutation‐carriers and 96 non‐carriers) to find groups of proteins associated with mutation status and disease stage as measured with estimated year to symptom onset (EYO), PiB‐PET, and clinical status. Linear‐mixed effect models and two‐part zero‐inflated negative binomial mixed models further evaluated the link between significant principal components (PCs) and WMH and CMH, respectively. Models accounted for age, APOE‐e4 status, sex, and education.ResultsThe first ten PCs explained >60% of the variance in NVU‐related protein levels (Figure 2A) with only PC5 (4.8%) associated with mutation*EYO, suggesting relationship with ADAD‐specific disease progression (***p<0.0001, Figure 2B). While asymptomatic PiB‐negative individuals were not significantly different from non‐carriers, asymptomatic PiB‐positive individuals had higher PC5 scores than non‐carriers and symptomatic individuals scored still higher (**p<0.005, ***p<0.0001, Figure 2C). The main proteins composing PC5 were: neurofilament light‐chain, angiopoietin‐2, SMOC1, LRP1, E‐selectin, and sTREM2, suggesting strong association with proinflammatory glial response from neuronal injury (Figure 2D). PC5 was also associated with WMH volumes (*p<0.05, Figure 3A) and with CMH count in participants with CMH (**p<0.005, Figure 3B).ConclusionsThis PCA approach indicates that some groups of NVU‐associated proteins are related with mutation status and disease stage, and that they are differentially associated with features of ADAD such as white matter integrity as measured by WMH volume. Weighted Correlation Network Analysis will be of interest to further understand these associations.Funding: K01AG080123, RF1‐AG044546, UF1AG032438

AbstractBackgroundCerebral small vessel disease (cSVD), as defined by neuroimaging characteristics such as white matter hyperintensities (WMHs), cerebral microhemorrhages (CMHs), and lacunar infarcts, is highly prevalent and has been associated with dementia risk and other clinical sequelae. Although risk factors for cSVD have been identified, little is known about the biological processes and molecular mediators that influence cSVD development and progression.MethodsWithin the Atherosclerosis Risk in Communities (ARIC) study, we used SomaScan Multiplexed Proteomic technology to relate 4,877 plasma proteins to concurrently measured MRI‐defined cSVD characteristics, including WMHs, CMHs, and lacunar infarcts, in late‐life (n=1508; mean age: 76). Candidate protein associations with WMH volume were replicated in the Cardiovascular Health Study (n=765; mean age: 73) and Baltimore Longitudinal Study of Aging (n=902; mean age: 66). Cohort‐specific effect estimates were pooled using a fixed‐effect, inverse variance weighted meta‐analysis. Within 4228 non‐demented older adults in ARIC, we examined the association of WMH‐associated proteins with eight‐year dementia risk (641 incident dementia cases) using multivariate Cox proportional hazard models. All multivariate regression models were adjusted for demographic, kidney function, and cardiovascular risk variables.ResultsThis proteome‐wide analysis of older adults identified 13 WMH‐associated plasma proteins involved in synaptic function, endothelial integrity, and angiogenesis (Figures 1a, 2a). Nine of these proteins replicated in one or more external cohort (Figures 1b, 2b) and two proteins, oligodendrocyte myelin glycoprotein (OMG) and neuronal pentraxin receptor (NPTXR), remained associated with late‐life WMH volume when measured nearly two decades earlier during midlife (Figures 1c, 2c). Eight of the 13 WMH‐associated proteins were associated with lacunar infarcts, 6 with subcortical CMHs, and 4 with lobar CMHs (Figures 1d, 2d). Although none of the genes coding for WMH‐associated proteins have been nominated by WMH GWAS, 7 of these genes have been previously linked to cardiometabolic traits known to influence cSVD, such as blood pressure and cholesterol (Figures 1e, 2e). Finally, 11 candidate proteins were associated with incident dementia, underscoring their clinical relevance (Figures 1f, 2f).ConclusionsThis large‐scale proteomic analysis within longitudinal community‐based samples identified several plasma proteins, including OMG and NPTXR, as top candidate biomarkers for elevated WMH volume and its clinical manifestations.

AbstractBackgroundWomen with suspected coronary microvascular dysfunction (CMD) may be at higher risk of experiencing cognitive decline due to cerebral small vessel disease, a known contributor to Alzheimer’s disease and related dementias (ADRD). A potential underlying mechanism that could accelerate this cognitive decline is the accumulation of brain tissue iron, which has been previously linked to changes in brain function potentially caused by oxidative stress and cell death. Therefore, we aim to elucidate whether a similar mechanism could affect women with suspected CMD by investigating the potential role of iron deposition on the brain’s functional organization and its effect on cognition using advanced magnetic resonance imaging (MRI).Material and MethodsTwenty‐seven women with suspected CMD [Age; median (range) = 54 (29‐76)], drawn from ongoing cohorts (3R01HL146158‐04S1,3U54AG065141‐04S1), underwent a 3T MRI protocol, including submillimeter T2* 3‐dimensional echo‐planar‐imaging for assessing iron deposition with high‐resolution quantitative susceptibility mapping (QSM) and resting‐state fMRI (rs‐fMRI). Iron content was quantified by total‐generalized‐variation based QSM analysis. Functional integrity was determined via graph theoretical approach (i.e., nodal degree). Cognitive assessment was also performed using the NIH Toolbox. Mediation analysis was conducted using Python Statsmodels.ResultsMost of the women with suspected CMD showed lower performance in processing speed, working memory, executive function, and attention (Figure 1A). We found a significant association between elevated iron levels in paracentral gyrus and lower functional connectivity in left hippocampus (p=0.005, adjusted‐r2=0.19) (Figure 1B). Elevated iron level in the paracentral gyrus showed an impact on cognitive performance in the domains of executive function and attention (p<0.0001, coefficient=531.09) as well as language functions and crystalized abilities (p=0.036, coefficient=362.45), mediated by functional connectivity in the left hippocampus (indirect effect on executive function / language: p = 0.034 / 0.038, coefficient = ‐254.54 /‐ 213.06) (Figure 1C).ConclusionsOur results suggest that changes in hippocampal functional organization are associated with cortical iron deposition and mediate its impact on cognitive performances. These changes may increase the risk of cognitive decline/ADRD or in women with suspected CMD. Future research in a larger cohort with a longitudinal design is necessary to validate and expand upon these findings.

AbstractBackgroundNormative models (NM) of brain metrics based on large, diverse populations offer novel strategies to detect individual brain abnormalities. To create an age‐dependent statistical model of brain microstructure over the human lifespan, we built the largest multi‐site NM of white matter (WM) diffusion tensor imaging (DTI) metrics based on 54,591 subjects. We used state‐of‐the‐art tools to adjust for site‐dependent effects.MethodWe used Hierarchical Bayesian Regression (HBR) to determine the age trajectory and lifespan centile curves (ages 3‐100 years) by merging data from different sections of the lifespan. We analyzed 19 international public datasets with the ENIGMA‐DTI protocol. Mean fractional anisotropy (FA), mean, axial, and radial diffusivity (MD, AxD, RD) were extracted for 21 bilateral ROIs from the JHU‐WM atlas and the whole WM skeleton. Regressions were run with each metric per ROI as a function of age and sex, and the DTI protocol as the batch effect (37 protocols). Z‐scores were derived for 81 patients with Alzheimer’s disease (AD; mean age: 77y±8.4, 46M/35F) and 225 MCI participants (mean age: 75.1y±8.1, 127M/98F) and extreme deviations (Z>|2|) were quantified for each condition. ROI‐wise areas under the ROC curve were calculated to determine the classification accuracy of the Z‐scores.ResultThe average FA for the overall WM skeleton reached its peak value at 27 years, with minima for MD, RD and AxD at 47, 42 and 54 years, respectively. We found extreme FA deviations (Z>2) for MCI (AUC=0.67) and AD (AUC=0.66) in the genu of the corpus callosum, extreme RD deviations (Z<2) in the tapetum (AUC=0.73) and Extreme AxD deviations (Z>2) for MCI in the superior longitudinal fasciculus (AUC=0.65). See Figures 1 & 2.ConclusionNM of brain microstructure with HBR yielded lifespan trajectories for widely‐used DTI metrics. By pooling large scale multi‐site dMRI to create a statistical reference, single subject deviations are detectable across the lifespan for AD and MCI. These NM will be publicly available to the research community.

AbstractBackgroundThe HEAD study focuses on collecting an extensive dataset from various tau‐PET tracers, aiming to establish robust anchor values, which are essential for harmonizing tau‐PET measurements. Here, we aim to showcase the capability of converting 3D tau‐PET images into a common scale using the Universal Tau‐PET Scale, Uniτ (tau), and to use these 3D images to subsequently obtain ROIs as needed.MethodsWe assessed 185 individuals across the aging and AD spectrum from the HEAD study, with [18F]Flortaucipir and [18F]MK‐6240 tau‐PET tracers. Tau‐PET SUVR images were standardized to a common 8mm FWHM, using the inferior cerebellar gray matter as the reference region. We generated Uniτ Tau‐PET 3D images using a single formula based on Meta‐temporal parameters in two steps: first, within‐tracer anchoring based on young (<25 years) and cognitively impaired individuals; second, between‐tracer anchoring using a piecewise transformation from [18F]Flortaucipir to [18F]MK‐6240, with smoothing at the piecewise inflection point. Subsequently, we extracted mean Uniτ values from these 3D images for key ROIs, including meta‐temporal, mesial, temporo‐parietal, frontal, and Braak stages III‐VI. Finally, we correlated Uniτ values across ROIs between the two tracers to evaluate the accuracy of estimates from the voxel‐wise transformation.ResultsThe original SUVR images present large differences between the two tau‐PET tracers (Figure 1). However, upon applying the Uniτ piecewise transformation with smoothing to all brain voxels, we were able to reasonably harmonize these images to the Uniτ scale, substantially reducing visual variability. Notably, the mean Uniτ values for the ROIs extracted from these harmonized 3D tau‐PET brain images demonstrated a high level of association between the two tracers (Figure 2). Furthermore, estimates generated from ROIs or extracted from our 3D parametric Uniτ model yielded identical estimates (Figure 3).ConclusionThe strong associations between tracers after directly harmonizing 3D images to Uniτ scale using the piecewise transformation with smoothing, underscore the effectiveness of the proposed method. This approach provides a reliable and standardized way to compare tau‐PET data across different tracers. Our results indicate the feasibility of harmonizing 3D tau‐PET images without relying on pre‐established ROIs, overcoming the limitation of restricting the analysis to only few brain regions.