Many research studies have shifted their focus towards longitudinal data analysis. However, the currently available statistical software have not welcomed this shift as they are still lacking support for longitudinal study design, multiple imaging covariates or generalized linear regression analysis. Here, we introduce a novel statistical software written in Matlab which can perform longitudinal generalized regression analysis with multiple imaging covariates. The biggest challenge encountered to incorporate complex statistical models and multiple imaging covariates is the required time and memory complexity. This has been dealt by utilizing data parallelism techniques through the Matlab parallel computing toolbox and the Matlab distributed computing server. The function that performs generalized linear regression supports binomial, normal, poisson, gamma and inverse gaussian response variable distributions and can accommodate any number of imaging variables in the regression model and repeated measurements for longitudinal study designs. To illustrate the voxel-wise generalized linear regression functionality, neuroimaging data ([18F]FDG PET, T1-MRI) were acquired for 219 individuals from the ADNI database. Demographic and MMSE scores were also obtained for the same individuals to be included in the regression models. T1 data were processed using the CIVET image processing pipeline while the PET data were processed with an established image processing pipeline. The statistical model included a logistic regression analysis to evaluate the contribution from the interaction of grey matter density and glucose metabolism for developing Alzheimer’s dementia in a cohort of MCI patients. Figure 1. shows the brain regions with highest statistical significance to increase the odds of developing Alzheimer’s dementia within 24 months from MCI. Reduced glucose metabolism in the temporal brain structures and the precuneus show significant contribution toward increasing the odds of developing AD, while the interaction of reducing glucose metabolism and reducing grey matter density in the superior gyrus of the temporal lobe increases the odds of developing AD. This novel software enables rapid prototyping and testing of sophisticated image based hypotheses particularly involving longitudinal data with multispectral neuroimaging resources expanding the existing methods for neuroimage analysis. T-statistical maps for the FDG (a), VBM (b) and FDG*VBM (c) in logistic regression model for Convertion to AD in 24 months from MCI.

Although individuals presenting abnormalities of both amyloid-β (Aβ) deposition and neuronal injury are especially vulnerable to disease progression, the regional association of these two pathologies as a determinant of cognitive decline remains unclear. Here, we tested, at voxel level, whether co-localized interactions between abnormal Aβ and brain hypometabolism determine subsequent cognitive decline in individuals with mild cognitive impairment (MCI). In order to test this framework, we assessed 309 amnestic MCI individuals form ADNI cohort who underwent [18F]florbetapir and [18F]FDG positron emission tomography (PET) at baseline and a general cognitive performance testing (mini-mental state examination (MMSE)) at baseline and at 2-year follow-up visits (Table 1). First, cut-off analysis at every voxel was performed for both ligands, contrasting cognitively normal (n=209) and Alzheimer’s disease (AD) (n=81) individuals, based on the best operating point of the receiver operating characteristic (ROC) curve. Second, parametrical maps of normality and abnormality at every voxel were generated for both ligands using the aforementioned cut-off maps for each MCI individual. Finally, using Voxel-stats, a voxel-based interaction model was built to assess the brain regions in which the coexistence of an abnormal [18F]florbetapir and [18F]FDG standardized uptake ratios (SUVR) interacted to determine an increased rate of cognitive decline over 2 years, as described in the formula: ΔMMSE = β0 + β1(florbetapir SUVR (0,1)) + β2(FDG SUVR(0,1)) + β3(florbetapir SUVR (0,1)*FDG SUVR (0,1)) + covariates + error. The statistical parametric maps were adjusted for age, gender, education, APOE ε4 status, baseline MMSE and corrected for multiple testing at a threshold of P < 0.05. The voxel-based interaction model revealed that a synergistic effect between abnormal [18F]florbetapir and [18F]FDG SUVRs in the precuneus and posterior cingulate cortices determined a faster rate of cognitive decline in amnestic MCI individuals (P < 0.05) (Figure 1). Our results support the notion that the regional synergism between brain Aβ and metabolic injury is associated with subsequent clinical progression in patients with MCI.

Conceptual framework suggests that the synergism between amyloid-β (Aβ) and phosphorylated tau (p-tau) aggregates determines structural and functional abnormalities in cognitively normal individuals. Our study was designed to test the hypothesis that clinical progression to Alzheimer’s disease dementia (AD) in patients with mild cognitive impairment (MCI) is dependent on the synergism between Aβ and p-tau rather than merely additive effects of overlapping brain pathologies. We categorized 310 ADNI participants with amnestic MCI in four biomarker groups based on elevated or non-elevated Aβ positron emission tomography (PET) using [18F]Florbetapir ligand and cerebrospinal fluid p-tau biomarkers (Table 1). Clinical and neuropsychological evaluations were performed at baseline and at 2 years including memory, executive function, psychomotor speed processing, and language. Regression interaction models evaluated changes in cognition and clinical status as a function of baseline imaging and fluid biomarkers. Using Voxel-Stats, we built a voxel-based logistic regression model in order to test the association between Aβ PET at every voxel and p-tau status as determinant of progression to dementia, assuming the probability of progression as p’: Log p’ / 1-p’ = β0 + β1(florbetapir SUVR) + β2(CSF p-tau status) + β3(florbetapir SUVR*p-tau status) + covariates + error. All models were adjusted for age, gender, education, APOE ε4 status and corrected for multiple comparison testing using Bonferroni at P < 0.05. We found that Aβ+/p-tau+ individuals had the highest rate of clinical and neuropsychological decline in all cognitive domains as compared to Aβ+/p-tau−, Aβ−/p-tau+ and Aβ−/p-tau−, which did not differ from one another. Notably, regression interaction models confirmed that the synergistic effect between Aβ and p-tau abnormalities best-predicted cognitive decline and clinical progression to dementia, as compared to the sum of their independent effects. Finally, voxel-based logistic regression analysis revealed that the lateral and basal temporal and inferior parietal cortices were the brain regions where the synergistic effect between Aβ and p-tau status determined an increased likelihood of progression to dementia (Figure 1). Together, the present results suggest that clinical progression to AD is driven by a synergistic rather than a merely additive effect between Aβ aggregation and tau hyperphosphorylation.

INTRODUCTION:Recent literature proposes that amyloid β (Aβ) and phosphorylated tau (p-tau) synergism accelerates biomarker abnormalities in controls. Yet, it remains to be answered whether this synergism is the driving force behind Alzheimer disease (AD) dementia. METHODS:We stratified 314 mild cognitive impairment individuals using [18F]florbetapir positron emission tomography Aβ imaging and cerebrospinal fluid p-tau. Regression and voxel-based logistic regression models with interaction terms evaluated 2-year changes in cognition and clinical status as a function of baseline biomarkers. RESULTS:We found that the synergism between [18F]florbetapir and p-tau, rather than their additive effects, was associated with the cognitive decline and progression to AD. Furthermore, voxel-based analysis revealed that temporal and inferior parietal were the regions where the synergism determined an increased likelihood of developing AD. DISCUSSION:Together, the present results support that progression to AD dementia is driven by the synergistic rather than a mere additive effect between Aβ and p-tau proteins.

In healthy individuals, behavioral outcomes are highly associated with the variability on brain regional structure or neurochemical phenotypes. Similarly, in the context of neurodegenerative conditions, neuroimaging reveals that cognitive decline is linked to the magnitude of atrophy, neurochemical declines, or concentrations of abnormal protein aggregates across brain regions. However, modeling the effects of multiple regional abnormalities as determinants of cognitive decline at the voxel level remains largely unexplored by multimodal imaging research, given the high computational cost of estimating regression models for every single voxel from various imaging modalities. VoxelStats is a voxel-wise computational framework to overcome these computational limitations and to perform statistical operations on multiple scalar variables and imaging modalities at the voxel level. VoxelStats package has been developed in Matlab® and supports imaging formats such as Nifti-1, ANALYZE and MINC v2. Prebuilt functions in VoxelStats enable the user to perform voxel-wise general and generalized linear models and mixed effect models with multiple volumetric covariates. Importantly, VoxelStats can recognize scalar values or image volumes as response variables and can accommodate volumetric statistical covariates as well as their interaction effects with other variables. Furthermore, this package includes built-in functionality to perform voxel-wise receiver operating characteristic analysis and paired and unpaired group contrast analysis. Validation of VoxelStats was conducted by comparing the linear regression functionality with existing toolboxes such as glim_image and RMINC. The validation results were identical to existing methods and the additional functionality was demonstrated by generating feature case assessments (t-statistics, odds ratio and true positive rate maps). In summary, VoxelStats expands the current methods for multimodal imaging analysis by allowing the estimation of advanced regional association metrics at the voxel level.

# ContentThis work is derived from the Alzheimer's Disease Neuroimaging Initiative 2 (ADNI2) and three samples from Montreal, Canada, as described in the following publication: Tam et al, Common effects of amnestic mild cognitive impairment on resting-state connectivity across four independent studies (dx.doi.org/10.3389/fnagi.2015.00242). It includes group brain parcellations for clusters generated from resting-state functional magnetic resonance images for 99 cognitively normal elderly persons and 129 patients with mild cognitive impairment. The parcellations have been generated using a method called bootstrap analysis of stable clusters (BASC, Bellec et al., 2010) and 8 resolutions of clusters (4, 6, 12, 22, 33, 65, 111, and 208 total bihemispheric parcels) were selected using a data-driven method called MSTEPS (Bellec, 2013). This work also includes parcellations that contain regions-of-interest (ROIs) that span only one hemisphere at 8 resolutions (10, 17, 30, 51, 77, 137, 199, and 322 total ROIs). It also includes maps illustrating uncorrected functional connectivity differences (t-maps) between patients and controls for four seeds/ROIs (superior medial frontal cortex, dorsomedial prefrontal cortex, striatum, middle temporal lobe). This release contains the following files: * README.md: a text description of the release. * brain_parcellation_mcinet_basc_(sym,asym)_(#)clusters.(mnc,nii).gz: 3D volumes (either in .mnc or .nii format) at 3 mm isotropic resolution, in the MNI non-linear 2009a space (http://www.bic.mni.mcgill.ca/ServicesAtlases/ICBM152NLin2009), at multiple resolutions of # clusters. Region number I is filled with Is (background is filled with 0s). Note that four versions of the templates are available, named with sym_mnc, asym_mnc, sym_nii or asym_nii. The mnc flavor contains files in the minc format, while the nii flavor has files in the nifti format. The asym flavor contains brain images that have been registered in the asymmetric version of the MNI brain template (reflecting that the brain is asymmetric), while with the sym flavor they have been registered in the symmetric version of the MNI template. The symmetric template has been forced to be symmetric anatomically, and is therefore ideally suited to study homotopic functional connections in fMRI: finding homotopic regions simply consists of flipping the x-axis of the template. Note: These clusters are often bihemispheric. For parcellations containing regions that span only one hemisphere, see below. * brain_parcellation_mcinet_basc_(sym,asym)_(#)rois.(mnc,nii).gz: 3D volumes (either in .mnc or .nii format) at 3 mm isotropic resolution, in the MNI non-linear 2009a space, at multiple resolutions of # ROIs, that span only one hemisphere. As above, mnc/nii and sym/asym versions of the templates are available.* labels_mcinet_(sym,asym)_ (#)(clusters,ROIs).csv: spreadsheets containing labels for each cluster or ROI * ttest_ctrlvsmci_seed(#).(mnc,nii).gz: 3D volumes (either in .mnc or .nii) displaying functional connectivity differences (uncorrected t-tests) between patients with mild cognitive impairment and cognitively normal elderly, for 4 different seeds/regions of interest i.e. striatum (seed #2), dorsomedial prefrontal cortex (#9), middle temporal lobe (#12), superior medial frontal cortex (#28); cluster numbers are taken from the parcellation containing 33 clusters. # Preprocessing The datasets were analysed using the NeuroImaging Analysis Kit (NIAK https://github.com/SIMEXP/niak) version 0.12.18, under CentOS version 6.3 with Octave (http://gnu.octave.org) version 3.8.1 and the Minc toolkit (http://www.bic.mni.mcgill.ca/ServicesSoftware/ServicesSoftwareMincToolKit) version 0.3.18. Each fMRI dataset was corrected for inter-slice difference in acquisition time and the parameters of a rigid-body motion were estimated for each time frame. Rigid-body motion was estimated within as well as between runs, using the median volume of the first run as a target. The median volume of one selected fMRI run for each subject was coregistered with a T1 individual scan using Minctracc (Collins and Evans, 1998), which was itself non-linearly transformed to the Montreal Neurological Institute (MNI) template (Fonov et al., 2011) using the CIVET pipeline (Ad-Dabbagh et al., 2006). The MNI symmetric template was generated from the ICBM152 sample of 152 young adults, after 40 iterations of non-linear coregistration. The rigid-body transform, fMRI-to-T1 transform and T1-to-stereotaxic transform were all combined, and the functional volumes were resampled in the MNI space at a 3 mm isotropic resolution. The “scrubbing” method of (Power et al., 2012), was used to remove the volumes with excessive motion (frame displacement greater than 0.5 mm). A minimum number of 50 unscrubbed volumes per run was then required for further analysis. The following nuisance parameters were regressed out from the time series at each voxel: slow time drifts (basis of discrete cosines with a 0.01 Hz high-pass cut-off), average signals in conservative masks of the white matter and the lateral ventricles as well as the first principal components (95% energy) of the six rigid-body motion parameters and their squares (Giove et al., 2009). The fMRI volumes were finally spatially smoothed with a 6 mm isotropic Gaussian blurring kernel. # Bootstrap Analysis of Stable Clusters Brain parcellations were derived using BASC (Bellec et al. 2010). A region growing algorithm was first applied to reduce the brain into regions of roughly equal size, set to 1000 mm3. The BASC used 100 replications of a hierarchical clustering with Ward's criterion on resampled individual time series, using circular block bootstrap. A consensus clustering (hierarchical with Ward's criterion) was generated across all the individual clustering replications pooled together, hence generating group clusters. The generation of group clusters was itself replicated by bootstrapping subjects 500 times, and a consensus clustering (hierarchical Ward's criterion) was generated on the replicated group clusters. The MSTEPS procedure (Bellec et al., 2013) was implemented to select a data-driven subset of scales in the range 5-200, approximating the group stability matrices up to 5% residual energy, through linear interpolation over selected scales. This resulted in 8 resolutions: 4, 6, 12, 22, 33, 65, 111, 208. Note that the number of resolutions was selected by the MSTEPS procedure in a data-driven fashion, and that the number of individual, group and final (consensus) number of clusters were not necessarily identical. These 8 resolutions were further processed to generate 8 parcellations that contain ROIs that span only one hemisphere. These latter parcellations contain 10, 17, 30, 51, 77, 137, 199, and 322 total ROIs. # Derivation of functional connectomes For each resolution K, and each pair of distinct clusters, the between-clusters connectivity was measured by the Fisher transform of the Pearson’s correlation between the average time series of the clusters. The within-cluster connectivity was the Fisher transform of the average correlation between time series inside the cluster. An individual connectome was thus a KxK matrix. # Statistical testing To test for differences between patients and controls at a given resolution, we used a general linear model (GLM) for each connection between two clusters. The GLM included an intercept, the age and sex of participants, and the average frame displacement of the runs involved in the analysis. The contrast of interest (patients–controls) was represented by a dummy covariate coding the difference in average connectivity between the two groups. All covariates except the intercept were corrected to a zero mean. The GLM was estimated independently for each scanning protocol. The estimated effects were combined across all protocols through inverse variance based weighted averaging (Willer et al., 2010). # References Ad-Dab’bagh, Y, et al, 2006. The CIVET Image-Processing Environment: A Fully Automated Comprehensive Pipeline for Anatomical Neuroimaging Research. In: Corbetta, M (Ed.), Proceedings of the 12th Annual Meeting of the Human Brain Mapping Organization. Neuroimage, Florence, Italy. Bellec, P, et al, 2010. Multi-level bootstrap analysis of stable clusters in resting-state fMRI. NeuroImage 51 (3), 1126–1139. Bellec, P, Jun. 2013. Mining the Hierarchy of Resting-State Brain Networks: Selection of Representative Clusters in a Multiscale Structure. In: Pattern Recognition in Neuroimaging (PRNI), 2013 International Workshop on. pp. 54–57. Collins, DL, Evans, AC, 1997. Animal: validation and applications of nonlinear registration-based segmentation. International Journal of Pattern Recognition and Artificial Intelligence 11, 1271–1294. Fonov, V, et al, 2011. Unbiased average age-appropriate atlases for pediatric studies. NeuroImage 54 (1), 313–327. Giove, F, et al, 2009. Images-based suppression of unwanted global signals in resting-state functional connectivity studies. Magnetic resonance imaging 27 (8), 1058–1064. Power, JD, et al, 2012. Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion. NeuroImage 59 (3), 2142–2154. Willer, CJ, et al, 2010. METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics 26, 2190–2191. # Other derivatives The datasets that were used to generate the parcellations are described in a publication, see the following link: https://github.com/SIMEXP/mcinet

During the last decades researchers have been using global measurements of amyloid−PET ligands to dichotomize subjects into amyloid-β (Aβ) positive or negative groups. The Aβ dichotomization is desirable to enrich clinical trials population and to assess the influences of Aβ abnormalities on Alzheimer’s disease (AD) progression. However, dichotomizations using global measurements do not provide information regarding the regional pattern of Aβ abnormalities, which may be important to identifying nondemented individuals fated to AD clinical progression. Here, we tested the framework that cut-off analysis performed at every voxel may provide additional information as compared to global estimates. We assessed cognitively normal (n=209), mild cognitive impairment (MCI; n=311) and AD (n=81) individuals from ADNI cohort who underwent [18F]Florbetapir PET at baseline (Table 1). The standardized uptake value ratio (SUVR) maps were then generated using the cerebellum grey matter and the global white matter as reference regions. First, a receiver operating characteristic (ROC) curve was performed at every voxel contrasting controls and AD participants. Second, the optimal cut-off value at every voxel was calculated using the least distance from (0,1) point to the ROC curve (best operating point) (Figure 1). Third, parametric maps of diagnostic sensitivity and specificity were generated (Figure 2). Finally, probabilistic maps for baseline Aβ positivity at every voxel were generated for MCI converters (n= 55) and non-converters (n= 256) over 2 years (Figure 3). [18F]Florbetapir SUVR cut-off values at every voxel. [18F]Florbetapir SUVR cut-off values sensitivity and specificity for a diagnostic of probable Alzheimer's disease at every voxel. Probabilistic maps of [18F]Florbetapir SUVR positivity at every voxel for MCI Non-converters and converters. The highest SUVR cut-off values were found in the precuneus, anterior and posterior cingulate cortices, whereas the lowest were found in clusters in the temporal lobe (Figure 1). Diagnostic sensitivity and specificity were the highest in clusters in the precuneus, posterior cingulate, temporal, and frontal cortices (Figure 2). Probabilistic maps showed that MCI non-converters did not present a specific pattern of amyloid deposition at baseline, whereas MCI converters reached 100% of positivity in voxels in the posterior cingulate, precuneus, frontal and temporal cortices (Figure 3). Our results revealed that the analysis of amyloid-PET cut-offs at every voxel might provide important information regarding the patterns of regional Aβ abnormalities associated with the clinical progression to AD.

Resting-state functional connectivity is a promising biomarker for Alzheimer's disease. However, previous resting-state functional magnetic resonance imaging studies in Alzheimer's disease and amnestic mild cognitive impairment (aMCI) have shown limited reproducibility as they have had small sample sizes and substantial variation in study protocol. We sought to identify functional brain networks and connections that could consistently discriminate normal aging from aMCI despite variations in scanner manufacturer, imaging protocol, and diagnostic procedure. We therefore combined four datasets collected independently, including 112 healthy controls and 143 patients with aMCI. We systematically tested multiple brain connections for associations with aMCI using a weighted average routinely used in meta-analyses. The largest effects involved the superior medial frontal cortex (including the anterior cingulate), dorsomedial prefrontal cortex, striatum, and middle temporal lobe. Compared with controls, patients with aMCI exhibited significantly decreased connectivity between default mode network nodes and between regions of the cortico-striatal-thalamic loop. Despite the heterogeneity of methods among the four datasets, we identified common aMCI-related connectivity changes with small to medium effect sizes and sample size estimates recommending a minimum of 140 to upwards of 600 total subjects to achieve adequate statistical power in the context of a multisite study with 5-10 scanning sites and about 10 subjects per group and per site. If our findings can be replicated and associated with other established biomarkers of Alzheimer's disease (e.g., amyloid and tau quantification), then these functional connections may be promising candidate biomarkers for Alzheimer's disease.

Effective self-control relies on the rapid adjustment of inappropriate responses. Understanding the brain basis of these processes has the potential to inform neurobiological models of the many neuropsychiatric disorders that are marked by maladaptive responding. Research on error processing in particular has implicated the dorsomedial frontal lobe (DMF) and basal ganglia (BG) in error detection, inhibition and correction. However there is controversy regarding the specific contributions of these regions to each of these component processes. Here we examined the effects of lesions affecting DMF or BG on these error-related processes. A flanker task was used to induce errors that in turn led to spontaneous, online corrections, while response kinematics were measured with high spatiotemporal resolution. The acceleration of errors was initially greater than that of correct responses. Errors then showed slower acceleration compared to correct responses, consistent with engagement of inhibition shortly after error response onset. BG damage disproportionately disrupted this early inhibitory phenomenon, above and beyond effects on baseline motor performance, but did not affect the kinematics of the corrective response. DMF damage showed the opposite pattern, with relatively delayed onset and weaker initial acceleration of the corrective response, but error suppression kinematics similar to that of the control group. This work clarifies the component processes and neural substrates of online post-error control, providing evidence for dissociable contributions of BG to error inhibition, but not correction, and DMF to rapid error correction, but not error suppression.

Accurate diagnosis of Alzheimer's disease and its prodromal state is of paramount importance for effective intervention. Recent studies have shown that imaging biomarkers provides excellent knowledge for classification but failed to account to a compelling classifier due to usage of consolidated information (global SUVR measurements). We hypothesize that using voxel-based information we would be able to better classify Alzheimer's individuals from cognitively normal individuals. [18F]Florbetapir PET images were acquired from 83 subjects (65 Cognitively Normal [CN], 18 Alzheimer's Disease [AD]) from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. The respective standardized uptake value ratio (SUVR) maps were subsequently generated using cerebellar grey-matter as reference region. Corresponding cortical uptake values (vertex based) were extracted using an average mid-surface structure generated using the study subjects and vertex based Receiver Operating Characteristic (ROC) analysis was carried out to identify the brain regions that best discriminates patients from cognitively normal individuals. ROC analysis based on the consolidated measurements was also carried out as a comparison study. Based on the Area Under the Curve (AUC) values, brain regions including precuneus, posterior cingulate cortex, medial orbitofrontal cortex and temporal lobe showed the best separation between patients and normal individuals with an AUC value of over 0.8. The same regions show sensitivity values of over 0.7 and specificity values of over 0.8 (Figure 1). In the study done using consolidated measures, the best separation resulted in AUC of 0.6872 with a specificity of 0.8 and sensitivity of 0.722. Image shows Area Under the Curve (AUC), sensitivity and specificity of the regional ROC analysis. Regions including the precuneus, posterior cingulate cortex, medial orbitofrontal cortex and temporal lobe showed the highest AUC value indicating the best separation between AD patients and CN individuals. These regions also show a high sensitivity and specificity. ROC estimate of regional concentrations of brain [18F]Florbetapir may contribute to the identification of pathological patterns of amyloidosis in predementia population and will enable accurate and effective classification in to the disease stages. The preliminary data reported here support the hypothesis that regional amyloidosis might contribute to a better discrimination between AD patients and cognitively normal individuals when compared with a global measurement, and the regions that allows best separation need to be used to generate a better consolidated measurement.

Figure 1. Vertex-based multivariate linear regression model showing the effect of amyloid load on the rate of hypo-metabolism in each disease stage, corrected for baseline glucose metabolism, age, gender and apoe genotype. Only LMCI and AD stages show positive effect from amyloid load on hypometabolism in temporo-parietal and precuneus regions. Sulantha S. Mathotaarachchi, Sara Mohades, Monica Shin, Thomas Beaudry, Andrea Lessa Benedet, Tharick Ali Pascoal, Seqian Wang, Sarinporn Manitsirikul, Maxime J. Parent, Min Su Kang, Vladimir Fonov, Chang Oh Chung, Sr., Serge Gauthier, Pedro RosaNeto, McGill University, Montreal, QC, Canada; McGill Centre for Studies in Aging, Montreal, QC, Canada; McGill Centre for Studies in Aging/Translational Neuroimaging Laboratory, Montreal, QC, Canada; McGill Centre for Studies in Aging, Verdun, QC, Canada; McGill University Centre for Studies in Aging, Verdun, QC, Canada; Image Processing Laboratory, Montreal Neurological Institute, McGill University, Montreal, QC, Canada; Centre for Studies on Prevention of Alzheimer’s Disease (StoP-AD Centre), Douglas Mental Health Institute, Montreal, QC, Canada; Douglas Hospital Research Centre, Montreal, QC, Canada; Translational Imaging Laboratory, Montreal, QC, Canada. Contact e-mail: sulantha.s@gmail.com

Global amyloid burden is widely used as measurement to assess the brain amyloidosis in many studies, but its reliability in a longitudinal study is yet to be fully understood. In this study we developed a surface based technique to assess the relationship between amyloidosis and neurodegeneration measured by hypo-metabolism across diagnosis stages of Alzheimer's disease. We hypothesize that the regional effects of amyloid retention on the rate of hypo-metabolism depend on the diagnosis stage. We will further investigate whether a consolidated measure of amyloidosis (global SUVR) can capture the effects shown in the regional analysis. The study included 213 subjects (65 Cognitively Normal [CN], 111 Early Cognitive Impairment [EMCI], 19 Late Cognitive Impairment [LMCI], 18 Alzheimer's Disease [AD]) taken from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Their [18F]Florbetapir and [18F]FDG PET images were acquired 24 months apart and the SUVR maps were subsequently generated using cerebellar grey/white matter and pons as reference regions for [18F]Florbetapir and [18F]FDG PET images, respectively. The effects of [18F]Florbetapir on yearly rate of metabolism were computed using vertex-based regression models including the baseline glucose metabolism, age, gender and APOE genotype as covariates. During the 24-month observation period, for the subjects in CN and EMCI stages, the regional rates of hypo-metabolism did not show a linear relationship to the localized amyloid burden. However, at LMCI and AD stages, regional amyloid load shows a linear relationship to the rate of metabolic decline in precuneus and temporo-parietal areas (Figure 1). In contrast, the global measure of amyloid burden showed a linear relationship with metabolism only in temporal regions. Vertex-based multivariate linear regression model showing the effect of amyloid load on the rate of hypo-metabolism in each disease stage, corrected for baseline glucose metabolism, age, gender and apoe genotype. Only LMCI and AD stages show positive effect from amyloid load on hypo-metabolism in temporo-parietal and precuneus regions. Based on the results it is evident that in LMCI and AD stages, the rates of hypo-metabolism at temporo-parietal areas as well as in the precuneus depend on the local tissue amyloid burden, while in CN and EMCI stages, it is independent. Absence of a relationship in similar regions with the global measure amyloidosis suggests that a consolidated measure cannot represent the brain amyloid burden in all disease stages, and perhaps, a new consolidation masks need to be generated based on disease stage.

Accurate diagnosis of Alzheimer's disease and its prodromal state is of paramount importance for effective intervention. Recent studies have shown that imaging biomarkers provides excellent knowledge for classification but failed to account to a compelling classifier due to usage of consolidated information (global SUVR measurements). We hypothesize that using voxel-based information we would be able to better classify Alzheimer's individuals from cognitively normal individuals. [18F]Florbetapir PET images were acquired from 83 subjects (65 Cognitively Normal [CN], 18 Alzheimer's Disease [AD]) from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. The respective standardized uptake value ratio (SUVR) maps were subsequently generated using cerebellar grey-matter as reference region. Corresponding cortical uptake values (vertex based) were extracted using an average mid-surface structure generated using the study subjects and vertex based Receiver Operating Characteristic (ROC) analysis was carried out to identify the brain regions that best discriminates patients from cognitively normal individuals. ROC analysis based on the consolidated measurements was also carried out as a comparison study. Based on the Area Under the Curve (AUC) values, brain regions including precuneus, posterior cingulate cortex, medial orbitofrontal cortex and temporal lobe showed the best separation between patients and normal individuals with an AUC value of over 0.8. The same regions show sensitivity values of over 0.7 and specificity values of over 0.8 (Figure 1). In the study done using consolidated measures, the best separation resulted in AUC of 0.6872 with a specificity of 0.8 and sensitivity of 0.722. Image shows Area Under the Curve (AUC), sensitivity and specificity of the regional ROC analysis. Regions including the precuneus, posterior cingulate cortex, medial orbitofrontal cortex and temporal lobe showed the highest AUC value indicating the best separation between AD patients and CN individuals. These regions also show a high sensitivity and specificity. ROC estimate of regional concentrations of brain [18F]Florbetapir may contribute to the identification of pathological patterns of amyloidosis in predementia population and will enable accurate and effective classification in to the disease stages. The preliminary data reported here support the hypothesis that regional amyloidosis might contribute to a better discrimination between AD patients and cognitively normal individuals when compared with a global measurement, and the regions that allows best separation need to be used to generate a better consolidated measurement.

Alzheimer's disease (AD) is a complex disease in which the quest for causative genes has been challenging. In this search, genome-wide studies (GWAS) have been a valuable tool, being able to investigate, through thousands of markers, associations with the disease or with its endophenotypes. However, GWAS analyses almost always use cross-sectional data. Despite the adversities of obtaining and analyzing longitudinal data, the information about the disease progress can be of great importance. For this reason, we have searched for genetic markers associated with changes in brain amyloid (Aβ) load and glucose uptake. [18F]Florbetapir positron emission tomography (PET) imaging was employed to assess brain Aβ levels in 412 participants from the Alzheimer's Disease Neuroimaging Initiative, whilst glucose uptake was measured using [18F]fludeoxyglucose PET (FDG) in 419 subjects from the same cohort. The genotypes were obtained with IlluminaHumanOmni2.5 beadchip. After quality control in both imaging and genetic data, a GWAS was performed. The phenotypes used were the differences between global SUVR in the baseline and 24 months follow-up of [18F]florbetapir and FDG. Covariates as diagnostic status and baseline SUVR were added in the genetic analysis. The Bonferroni threshold of genome-wide significance is 3.9x10−8. Values higher then 3.9x10−8 but less then 10−6were considered trends of association. None of the SNPs reached genome-wide significance, however trends of association are reported here (Figure 1). Aβ accumulation shows a trend with 24 markers from 15 genes, in which probably the most relevant, and significant, is PLCH1. Brain hipometabolism indicate trend associations with 21 markers from 8 genes. The genes TTC39B and NRP1were the most significant. A) Quantile-quantile plot of [l8F]florbetabir GWAS. B) Manhattan plot [l8F]florbetabir GWAS. Red line represents the trend threshold of significance. C) Quantile-quantile plot of FDG GWAS. D) Manhattan plot FDG GWAS. Red line represents the trend threshold of significance. The major genes reported to be associated with AD were not found in the present study. Interestingly, Aβ accumulation seems to be related to the PLCH1 gene, which encodes for a phospholipase-C family-member. Phospholipases are important enzymes with key roles in cell signaling. Another relevant result may be the association of hypometabolism with the TTC39B gene. Its function is not clear, however this gene has been related to lipid metabolism. Further studies with bigger sample size would be necessary to confirm present results.

Resting-state functional connectivity is a promising biomarker for Alzheimer’s disease. However, previous resting-state functional magnetic resonance imaging studies in Alzheimer’s disease and mild cognitive impairment (MCI) have shown limited reproducibility as they have had small sample sizes and substantial variation in study protocol. We sought to identify functional brain networks and connections that could consistently discriminate normal aging from MCI despite variations in scanner manufacturer, imaging protocol, and diagnostic procedure. We therefore pooled four independent datasets, including 112 healthy controls and 143 patients with MCI, systematically testing multiple brain connections for consistent differences. The largest effects associated with MCI involved the ventromedial and dorsomedial prefrontal cortex, striatum, and middle temporal lobe. Compared with controls, patients with MCI exhibited significantly decreased connectivity within the frontal lobe, between frontal and temporal areas, and between regions of the cortico-striatal-thalamic loop. Despite the heterogeneity of methods among the four datasets, we identified robust MCI-related connectivity changes with small to medium effect sizes and sample size estimates recommending a minimum of 150 to 400 total subjects to achieve adequate statistical power. If our findings can be replicated and associated with other established biomarkers of Alzheimer’s disease (e.g. amyloid and tau quantification), then these functional connections may be promising candidate biomarkers for Alzheimer’s disease.

Identification of cognitively normal (CN) individuals destined to develop hypometabolism has immediate applications on preventive clinical trials. Current biomarker progression models of Alzheimer's disease propose abnormal tau hyperphosphorylation preceding hypometabolism. From the Alzheimer's Disease Neuroimaging Initiative (ADNI), we investigated whether combined measures of amyloid and tau pathology predict subsequent hypometabolism in cognitively normal subjects. CN individuals (n=108) who had CSF total tau (t-tau), phospho-tau181p (p-tau), Aβ1–42, [18F]FDG and [18F]Florbetapir measurements in baseline and 24 month follow-up were analyzed. Subjects were dichotomized using CSF Aβ1–42, t-tau and p-tau published cutoffs, and further divided in [18F]Florbetapir positive and negative. [18F]FDG and [18F]Florbetapir PET SUVRs were computed using the pons and cerebellum as reference regions, respectively. We tested our hypothesis using voxel-based linear model, which included 24 month [18F]FDG PET SUVR changes as a response variable and global [18F]Florbetapir PET SUVR, CSF Aβ1–42, t-tau, p-tau status as factors; model was corrected for age, gender and APOE4 status. Results were FDR corrected (p < 0.001). Table 1 summarizes the study demographics. Figure 1 shows areas in which metabolic declines in [18F]FDG were predicted by the interaction of baseline p-tau and [18F]Florbetapir PET status. [18F]FDG SUVRs did not differ between the groups composed by the dichotomization of CSF p-tau and Global [18F]Florbetapir PET. In contrast, the group with abnormal CSF p-tau and [18F]Florbetapir PET status had 20-30% lower follow-up [18F]FDG SUVRs mean in the mesial temporal structures. Voxel-based factorial using CSF Aβ1–42 or t-tau values showed no significant results.

Alzheimer's disease (AD) is characterized by numerous pathological features, including amyloidosis, neurofibrillary tangles (NFTs) and brain atrophy. The specific contribution of NFTs and amyloid plaques toward the severity of AD has been controversial. While many animal models of AD show reduction in the hippocampal volume, it is unknown whether the atrophy is due to amyloidosis or NFTs. Our pilot data showed hippocampal volume reduction in animals older than 18 months. We aim to investigate if amyloid plaques alone are sufficient to trigger hippocampus-specific atrophy in the brains of McGill-R-Thy1-APP transgenic (TG) rats in comparison with wild-type (WT). Whole brain image acquisition consisted of Bruker 70/30USR Biospect MRI using a Fast Imaging with Steady State Precession for a total scan time just over 45 minutes (FISP; TE/TR: 2.5/5.0ms; FOV: 3.6 x 3.6 x 3.6 cm; isotropic 250um voxels; 8 angles). The cohort consisted of WT (N=6) and TG (N=10) rats which had MRI scans under 2% Isoflurane anesthesia at 11-month of age for baseline and 18-month as follow-up. Hippocampal 3D manual volumetry was performed using the software MINC- Display. Coronal images of commissural fornix and amygdala set the anterior hippocampal landmarks. Segmentation continued posteriorly until the granular layer and CA1 disappeared. Ventricles and the mesencephalic cistern served as the medial and lateral landmarks, respectively. Subiculum set the inferior and posterior hippocampal limits. Adjustments were guided by the correspondent sagittal and lateral views of the brains. WT showed 5.13% atrophy of the hippocampal formation, from 80.45±5.61mm 3 to 76.07±2.44mm 3, while TG showed 6.15%, from 80.46±8.90mm 3 to 75.48±8.74mm 3 (P<0.001). When normalized for the intracranial volumes, the extent of atrophy increased to 6.44% and 8.60% for both WT and TG, respectively (P<0.05, 0.01). A small but not significant group difference was apparent when comparing WT versus TG rats in the seven-month follow-up. Correlation of these data with cognitive assessments was also non significant.

It has been accepted that amyloidosis and hypometabolism propagates over time and accumulate across brain regions in the Alzheimer's Disease (AD), however patterns of propagation remains debatable. Modeling structural and functional brain regional associations allows for examining patterns of propagation and accumulations of AD pathophysiological processes in the living human brain. Dynamics of large-scale organization might serve as a sensitive metric to quantify patterns of amyloidosis or degenerative changes present in individuals at risk of developing AD dementia. Twenty four month longitudinal structure MRIs, [18F]FDG and [18F]Florbetapir PET images obtained from 206 subjects (60 Healthy Controls, 108 Early Cognitive Impairment, 19 Late Cognitive Impairment, 19 Alzheimer's Disease) were downloaded from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Surfaces were acquired using CIVET. The respective standardized uptake value ratio (SUVR) maps were subsequently generated using the median counts at cerebellar grey matter and pons for [18F]Florbetapir and [18F]FDG PET images, respectively. The SUVR at the cortex were projected using the mid brain surface of each subject. Vertex values were down sampled using k-means clustering algorithm. Regional Association Maps (RAMs) were then generated by using the correlation matrix and Bonferroni corrected regional association density maps were calculated for [18F]FDG and [18F]florbetapir. Comparing the baseline and follow-up regional associations, Amyloid accumulation ([18F]Florbetapir SUVR) shows a decrease over time in all the disease stages. A decrease in regional association can also be observed in Amyloid accumulation with the disease severity. Hypometabolism ([18F]FDG SUVR) shows an opposite variation (increase) when comparing the baseline and follow-up regional associations but decreases with disease severity. However, the regions that shows association in their rate of Amyloid accumulation and hypometabolism decrease with the disease severity [Figure 1]. Regional Association Maps capture the dynamic architecture of amyloid and hypometabolism propagation, occurring as the disease progresses. Regional associations decline for Amyloid accumulation as a function of the disease progression can be attributed to the non-linear rate of amyloidosis (reaching a plateau). This behavior is also supported by the reduction in regional associations for the rate of amyloid accumulation with the increase in disease severity. Longitudinal variation of regional associativity.

Figure 1. Coe Voxel-based co on 16 differen A NOVEL FRAMEWORK TO TEST THE DIAGNOSTIC USE OF DTI IN ALZHEIMER’S DISEASE Stefan Teipel, Giovanni B. Frisoni, Martin Dyrba, Andreas Fellgiebel, Massimo Filippi, Harald Hampel, University Medicine Rostock and DZNE Rostock, Rostock, Germany; IRCCS Centro San Giovanni di Dio, Fatebenefratelli, Brescia, Italy; German Center for Neurodegenerative Diseases (DZNE), Rostock/Greifswald, Rostock, Germany; University Medical Center Mainz, Mainz, Germany; Hospital San Raffaele, Milan, Italy; Universit e Pierre et Marie Curie, Paris, France. Contact e-mail: stefan.teipel@med.uni-rostock.de