Biology subjects

Alzheimer's Disease Neuroimaging Initiative,

Publications and source records attributed to Alzheimer's Disease Neuroimaging Initiative,.

At least 37 records · Page 2Linked to original sources

bioRxiv2024

The long distance relationship of regional amyloid burden and tau pathology spread

IntroductionConsistent with the amyloid-cascade-hypothesis, we tested whether regional amyloid burden is associated with tau pathology increases in spatially independent brain regions and whether functional connectivity serves as a mediator bridging the observed spatial gap between these pathologies. MethodsData of 98 amyloid-positive and 35 amyloid-negative subjects with baseline amyloid (18F-AV45) and longitudinal tau (18F-AV1451) PET were selected from ADNI. Annual tau change maps were computed. All images were z-transformed using the amyloid-negative subjects as reference. Z-maps of baseline amyloid and annual tau change were submitted to a parallel independent component analysis in GIFT, yielding six component pairs linking spatial patterns of baseline amyloid to longitudinal tau increase. Next, we used the region of maximum coefficient per component as seeds for functional connectivity analyses in a healthy control dataset. This resulted in six pairs of amyloid and tau seed-based networks (SBN). The spatial overlap between these SBNs and components (amyloid OR tau change) and the combined component pairs (amyloid AND tau change) were quantified. ResultsAmyloid SBNs presented greater spatial overlap with their respective amyloid components (24%-54%) than tau SBNs with the respective tau change components (16%-40%). However, the spatial combination of amyloid and tau component pairs showed highest spatial overlap with the amyloid SBNs (up to 62% vs. 39% for the tau SBNs). ConclusionMechanistically, regional associations of amyloid and tau pathology may be driven by underlying large-scale functional networks. Functional connections may thereby transmit soluble amyloid to remote brain regions within the same network, likely triggering tau aggregation.

neuroscience↗

bioRxiv2024

A novel classification framework for genome-wide association study of whole brain MRI images using deep learning

Genome-wide association studies (GWASs) have been widely applied in the neuroimaging field to discover genetic variants associated with brain-related traits. So far, almost all GWASs conducted in neuroimaging genetics are performed on univariate quantitative features summarized from brain images. On the other hand, powerful deep learning technologies have dramatically improved our ability to classify images. In this study, we proposed and implemented a novel machine learning strategy for systematically identifying genetic variants that lead to detectable nuances on Magnetic Resonance Images (MRI). For a specific single nucleotide polymorphism (SNP), if MRI images labeled by genotypes of this SNP can be reliably distinguished using machine learning, we then hypothesized that this SNP is likely to be associated with brain anatomy or function which is manifested in MRI brain images. We applied this strategy to a catalog of MRI image and genotype data collected by the Alzheimers Disease Neuroimaging Initiative (ADNI) consortium. From the results, we identified novel variants that show strong association to brain phenotypes.

bioinformatics↗

bioRxiv2023

Personalised Regional Modelling Predicts Tau Progression in the Human Brain

Aggregation of the hyperphosphorylated tau protein is a central driver of Alzheimers disease, and its accumulation exhibits a rich spatio-temporal pattern that unfolds during the course of the disease, sequentially progressing through the brain across axonal connections. It is unclear how this spatio-temporal process is orchestrated - namely, to what extent the spread of pathologic tau is governed by transport between brain regions, local production or both. To address this, we develop a mechanistic model from tau PET data to describe tau dynamics along the Alzheimers disease timeline. Our analysis reveals longitudinal changes in production and transport dynamics on two independent cohorts, with subjects in early stage of the disease exhibiting transport-dominated spread, consistent with an initial spread of pathologic tau seeds, and subjects in late stage disease (Braak stage 3/4 onwards) characterised primarily by local production of tau. Furthermore, we demonstrate that the model can accurately predict subject-specific longitudinal tau accumulation at a regional level, potentially providing a new clinical tool to monitor and classify patient disease progression. TeaserA mechanistic model reveals tau protein dynamics in Alzheimers, showing stage-specific shifts in transport and local production.

neuroscience↗

bioRxiv2023

Serum levels of adipokines and insulin are associated with markers of brain atrophy and cognitive decline in the spectrum of Alzheimer's Disease

The discovery that metabolic alterations often coexist with neurodegenerative conditions has sparked interest in the examination of gastrointestinal factors as potential modulators of brain health. Here, we examined the role of adipokines (leptin, adiponectin, resistin, and IL6) and insulin on different markers of brain atrophy in participants on the spectrum of Alzheimers Disease. We included 566 participants from the Alzheimers Disease Neuroimaging Initiative (ADNI) dataset with 1063 follow-up time points (average follow-up: one year); and examined the association between gastrointestinal factors and volumetric MRI values, white matter hyperintensities, and measures of cognitive impairment. Higher leptin, resistin, IL6, and insulin were associated with markers of cerebral atrophy, such as lower total brain volume, or higher ventricular volume. Higher leptin and resistin were also associated with greater impairment in daily life activities. Higher adiponectin was associated with lower ventricle volume. There was no association between adipokines or insulin with white matter hyperintensities. Our findings indicate a co-occurrence between alterations in gastrointestinal factors and in brain volume along the preclinical to clinical spectrum of Alzheimers Disease. These results suggest that strategies aimed at promoting metabolic health may positively impact brain health.

neuroscience↗

bioRxiv2023

Pathological brain states in Alzheimer's disease

Dynamic and rapid reconfigurations of neural activation patterns, known as brain states, support cognition. Recent analytic advances applied to functional magnetic resonance imaging now enable the quantification of brain states, which offers a substantial methodological improvement in characterizing spatiotemporal dynamics of activation over previous functional connectivity methods. Dysfunction to the persistence and temporal transitions between discrete brain states may be proximal factors reflecting neurophysiological disruptions in Alzheimers disease, although this has not yet been established. Here, we identified six distinct brain states, representing spatiotemporal trajectories of coactivation at single time points, in older adults across the Alzheimers disease continuum. Critically, we identified a pathological brain state that reflects coactivation within limbic regions. Higher persistence within and transitions to this limbic state, at the expense of other brain states, is associated with an increased likelihood of a clinically impaired diagnosis, worse cognitive performance, greater Alzheimers pathology, and neurodegeneration. Together, our results provide compelling evidence that neural activity settling into a pathological limbic state reflects the progression to Alzheimers disease. As brain states have recently been shown to be modifiable targets, this work may inform the development of novel neuromodulation techniques to reduce limbic state persistence. This application would be an innovative clinical approach to rescue cognitive decline in the early stages of Alzheimers disease.

neuroscience↗

bioRxiv2023

Exploratory Correlation of The Human Structural Connectome with Non-MRI Variables

INTRODUCTIONDiscovery of the associations between brain structural connectivity and clinical and demographic variables can help to better understand the vulnerability and resilience of the brain architecture to neurodegenerative diseases and to discover biomarkers. METHODSWe used four diffusion-MRI databases, three related to Alzheimers disease, to exploratorily correlate structural connections between 85 brain regions with non-MRI variables, while stringently correcting the significance values for multiple testing and ruling out spurious correlations via careful visual inspection. We repeated the analysis with brain connectivity augmented with multi-synaptic neural pathways. RESULTSWe found 85 and 101 significant relationships with direct and augmented connectivity, respectively, which were generally stronger for the latter. Age was consistently linked to decreased connectivity, and healthier clinical scores were generally linked to increased connectivity. DISCUSSIONOur findings help to elucidate which structural brain networks are affected in Alzheimers disease and aging and highlight the importance of including indirect connections.

neuroscience↗

bioRxiv2023

Leveraging longitudinal diffusion MRI data to quantify differences in white matter microstructural decline in normal and abnormal aging

AbstractO_ST_ABSINTRODUCTIONC_ST_ABSIt is unclear how rates of white matter microstructural decline differ between normal aging and abnormal aging. METHODSDiffusion MRI data from several well-established longitudinal cohorts of aging [Alzheimers Neuroimaging Initiative (ADNI), Baltimore Longitudinal Study of Aging (BLSA), Vanderbilt Memory & Aging Project (VMAP)] was free-water corrected and harmonized. This dataset included 1,723 participants (age at baseline: 72.8{+/-}8.87 years, 49.5% male) and 4,605 imaging sessions (follow-up time: 2.97{+/-}2.09 years, follow-up range: 1-13 years, mean number of visits: 4.42{+/-}1.98). Differences in white matter microstructural decline in normal and abnormal agers was assessed. RESULTSWhile we found global decline in white matter in normal/abnormal aging, we found that several white matter tracts (e.g., cingulum bundle) were vulnerable to abnormal aging. CONCLUSIONSThere is a prevalent role of white matter microstructural decline in aging, and future large-scale studies in this area may further refine our understanding of the underlying neurodegenerative processes. HIGHLIGHTSO_LILongitudinal data was free-water corrected and harmonized C_LIO_LIGlobal effects of white matter decline were seen in normal and abnormal aging C_LIO_LIThe free-water metric was most vulnerable to abnormal aging C_LIO_LICingulum free-water was the most vulnerable to abnormal aging C_LI

neuroscience↗

bioRxiv2022

Style Transfer Generative Adversarial Networks to Harmonize Multi-Site MRI to a Single Reference Image to Avoid Over-Correction

Recent work within neuroimaging consortia have aimed to identify reproducible, and often subtle, brain signatures of psychiatric or neurological conditions. To allow for high-powered brain imaging analyses, it is often necessary to pool MR images that were acquired with different protocols across multiple scanners. Current retrospective harmonization techniques have shown promise in removing cross-site image variation. However, most statistical approaches may over-correct for technical, scanning-related, variation as they cannot distinguish between confounded image-acquisition based variability and cross-site population variability. Such statistical methods often require that datasets contain subjects or patient groups with similar clinical or demographic information to isolate the acquisition-based variability. To overcome this limitation, we consider cross-site MRI image harmonization as a style transfer problem rather than a domain transfer problem. Using a fully unsupervised deep-learning framework based on a generative adversarial network (GAN), we show that MR images can be harmonized by inserting the style information encoded from a single reference image, without knowing their site/scanner labels a priori. We trained our model using data from five large-scale multi-site datasets with varied demographics. Results demonstrated that our style-encoding model can harmonize MR images, and match intensity profiles, without relying on traveling subjects. This model also avoids the need to control for clinical, diagnostic, or demographic information. We highlight the effectiveness of our method for clinical research by comparing extracted cortical and subcortical features, brain-age estimates, and case-control effect sizes before and after the harmonization. We showed that our harmonization removed the cross-site variances, while preserving the anatomical information and clinical meaningful patterns. We further demonstrated that with a diverse training set, our method successfully harmonized MR images collected from unseen scanners and protocols, suggesting a promising novel tool for ongoing collaborative studies. Source code is released in USC-IGC/style_transfer_harmonization (github.com).

neuroscience↗

bioRxiv2022

S-EBM: Generalising event-based modelling of disease progression for simultaneous events

This study introduces the parsimonious event-based model of disease progression (P-EBM). The P-EBM generalises the event-based model of disease progression (EBM) to allow inference of fewer disease progression stages than the number of input biomarkers. The original EBM is designed to estimate a single distinct biomarker abnormality, termed an event, at each model stage. By allowing multiple events within a common stage, the P-EBM prevents redundant complexity to permit discovery of parsimonious sequences of disease progression - those that contain purely serial events, as in the original EBM, as well as those containing one or more sets of simultaneous events. This study describes P-EBM theory, evaluates its sequence estimation and staging performance and demonstrates its application to derive a parsimonious sequence of disease progression in sporadic Alzheimers disease (AD). Results show that the P-EBM can accurately recover a wider range of sequences than EBM under a range of realistic experimental scenarios, including different numbers of simultaneous events, biomarker disease signals and dataset sizes. The P-EBM sequence successfully highlights redundant biomarkers and stages subjects using fewer biomarkers. In sporadic AD, the P-EBM estimates a shorter sequence than the EBM with substantially higher likelihood which plausibly suggests that some biomarker events appear simultaneously. The P-EBM has potential application for generating new insights into disease evolution and for suggesting efficient biomarker collection strategies for patient staging.

neuroscience↗

bioRxiv2022

Reproducibility of neuroimaging studies of brain disorders with hundreds -not thousands- of participants

An important current question in neuroimaging concerns the sample sizes required for producing reliable and reproducible results. Recent findings suggest that brain-wide association studies (BWAS) linking neuroimaging features with behavioural phenotypes in the general population are characterised by (very) weak effects and consequently need large samples sizes of 3000+ to lead to reproducible findings. A second, important goal in neuroimaging is to study brain structure and function under disease conditions, where effects are likely much larger. This difference in effect size is important. We show by means of power calculations and empirical analysis that neuroimaging studies in clinical populations need hundreds -and not necessarily thousands-of participants to lead to reproducible findings.

neuroscience↗

bioRxiv2022

Stage dependent differential influence of metabolic and structural networks on memory across Alzheimer's disease continuum

BackgroundLarge-scale neuronal network breakdown underlies memory impairment in Alzheimers disease (AD). However, the differential trajectories of the relationships between network organization and memory across pathology and cognitive stages in AD remain elusive. We determined whether and how the influences of individual-level structural and metabolic covariance network integrity on memory varied with amyloid pathology across clinical stages without assuming a constant relationship. Methods708 participants from the Alzheimers Disease Neuroimaging Initiative were studied. Individual-level structural and metabolic covariance scores in higher-level cognitive and hippocampal networks were derived from magnetic resonance imaging and [18F]fluorodeoxyglucose positron emission tomography using seed-based partial least square analyses. The non-linear associations between network scores and memory across cognitive stages in each pathology group were examined using sparse varying coefficient modelling. ResultsWe showed that the associations of memory with structural and metabolic networks in the hippocampal and default mode regions exhibited pathology-dependent differential trajectories across cognitive stages using sparse varying coefficient modelling. In amyloid pathology group, there was an early influence of hippocampal structural network deterioration on memory impairment in the preclinical stage, and a biphasic influence of the angular gyrus-seeded default mode network metabolism on memory in both preclinical and dementia stages. In non-amyloid pathology groups, in contrast, the trajectory of the hippocampus-memory association was opposite and weaker overall, while no metabolism covariance networks were related to memory. Key findings were replicated in a larger cohort of 1280 participants. ConclusionsOur findings highlight potential windows of early intervention targeting network breakdown at the preclinical AD stage.

neuroscience↗

bioRxiv2022

Hippocampal GFAP-positive astrocyte responses to amyloid and tau pathologies

IntroductionIn Alzheimers disease clinical research, glial fibrillary acidic protein (GFAP) released into the cerebrospinal fluid and blood is widely measured and perceived as a biomarker of reactive astrogliosis. However, it was demonstrated that GFAP levels differ in individuals presenting with amyloid-{beta} (A{beta}) or tau pathology. The molecular underpinnings behind this specificity are unexplored. Here we investigated biomarker and transcriptomic associations of GFAP-positive astrocytes with A{beta} and tau pathologies in humans and mouse models. MethodsWe studied 90 individuals with plasma GFAP, A{beta}- and Tau-PET to investigate the association between biomarkers. Then, transcriptomic analysis in hippocampal GFAP-positive astrocytes isolated from mouse models presenting A{beta} (PS2APP) or tau (P301S) pathologies was applied to explore differentially expressed genes (DEGs), Gene Ontology processes, and protein-protein interaction networks associated with each phenotype. ResultsIn humans, we found that plasma GFAP associates with A{beta} but not tau pathology. Unveiling the unique nature of GFAP-positive astrocytic responses to A{beta} or tau pathology, mouse transcriptomics showed scarce overlap of DEGs between the A{beta} and tau mouse models, While A{beta} GFAP-positive astrocytes were overrepresented with genes associated with proteostasis and exocytosis-related processes, tau hippocampal GFAP-positive astrocytes presented greater abnormalities in functions related to DNA/RNA processing and cytoskeleton dynamics. ConclusionOur results offer insights into A{beta}- and tau-driven specific signatures in GFAP-positive astrocytes. Characterizing how different underlying pathologies distinctly influence astrocyte responses is critical for the biological interpretation of astrocyte-related biomarker and suggests the need to develop context-specific astrocyte targets to study AD. FundingThis study was supported by Instituto Serrapilheira, Alzheimers Association, CAPES, CNPq and FAPERGS.

neuroscience↗

bioRxiv2021

Differentiating amyloid beta spread in autosomal dominant and sporadic Alzheimer's disease

Amyloid-beta (A{beta}) deposition is one of the hallmark pathologies in both sporadic Alzheimers disease (sAD) and autosomal dominant Alzheimers disease (ADAD), the latter of which is caused by mutations in genes involved in A{beta} processing. Despite A{beta} deposition being a centerpiece to both sAD and ADAD, some differences between these AD subtypes have been observed with respect to the spatial pattern of A{beta}. Previous work has shown that the spatial pattern of A{beta} in individuals spanning the sAD spectrum can be reproduced with high accuracy using an epidemic spreading model (ESM), which simulates the diffusion of A{beta} across neuronal connections and is constrained by individual rates of A{beta} production and clearance. However, it has not been investigated whether A{beta} deposition in the rarer ADAD can be modeled in the same way, and if so, how congruent the spreading patterns of A{beta} across sAD and ADAD are. We leverage the ESM as a data-driven approach to probe individual-level variation in the spreading patterns of A{beta} across three different large-scale imaging datasets (2 SAD, 1 ADAD). We applied the ESM separately to the Alzheimers Disease Neuroimaging initiative (N=737), the Open Access Series of Imaging Studies (N=510), and the Dominantly Inherited Alzheimers Network (N=249), the latter two of which were processed using an identical pipeline. We assessed inter- and intra-individual model performance in each dataset separately, and further identified the most likely epicenter of A{beta} spread for each individual. Using epicenters defined in previous work in sAD, the ESM provided moderate prediction of the regional pattern of A{beta} deposition across all three datasets. We further find that, while the most likely epicenter for most A{beta}-positive subjects overlaps with the default mode network, 13% of ADAD individuals were best characterized by a striatal origin of A{beta} spread. These subjects were also distinguished by being younger than ADAD subjects with a DMN A{beta} origin, despite having a similar estimated age of symptom onset. Together, our results suggest that most ADAD patients express A{beta} spreading patters similar to those of sAD, but that there may be a subset of ADAD patients with a separate, striatal phenotype.

neuroscience↗

bioRxiv2021

Multi-scale brain simulation with integrated positron emission tomography yields hidden local field potential activity that augments machine-learning classification of Alzheimer's disease

INTRODUCTIONComputational brain network modeling using The Virtual Brain (TVB) simulation platform acts synergistically with machine learning and multi-modal neuroimaging to reveal mechanisms and improve diagnostics in Alzheimers disease. METHODSWe enhance large-scale whole-brain simulation in TVB with a cause-and-effect model linking local Amyloid {beta} PET with altered excitability. We use PET and MRI data from 33 participants of Alzheimers Disease Neuroimaging Initiative (ADNI3) combined with frequency compositions of TVB-simulated local field potentials (LFP) for machine-learning classification. RESULTSThe combination of empirical neuroimaging features and simulated LFPs significantly outperformed the classification accuracy of empirical data alone by about 10% (weighted F1-score empirical 64.34% vs. combined 74.28%). Informative features showed high biological plausibility regarding the Alzheimers-typical spatial distribution. DISCUSSIONThe cause-and-effect implementation of local hyperexcitation caused by Amyloid {beta} can improve the machine-learning-driven classification of Alzheimers and demonstrates TVBs ability to decode information in empirical data employing connectivity-based brain simulation. RESEARCH IN CONTEXTO_LISYSTEMATIC REVIEW. Machine-learning has been proven to augment diagnostics of dementia in several ways. Imaging-based approaches enable early diagnostic predictions. However, individual projections of long-term outcome as well as differential diagnosis remain difficult, as the mechanisms behind the used classifying features often remain unclear. Mechanistic whole-brain models in synergy with powerful machine learning aim to close this gap. C_LIO_LIINTERPRETATION. Our work demonstrates that multi-scale brain simulations considering Amyloid {beta} distributions and cause-and-effect regulatory cascades reveal hidden electrophysiological processes that are not readily accessible through measurements in humans. We demonstrate that these simulation-inferred features hold the potential to improve diagnostic classification of Alzheimers disease. C_LIO_LIFUTURE DIRECTIONS. The simulation-based classification model needs to be tested for clinical usability in a larger cohort with an independent test set, either with another imaging database or a prospective study to assess its capability for long-term disease trajectories. C_LI

neuroscience↗

bioRxiv2020

Serum Metabolites Associate with Brain Amyloid Beta Deposition, Cognitive Dysfunction, and Alzheimers Disease Progression

RATIONALEMetabolomics in the Alzheimers Disease Neuroimaging Initiative (ADNI) cohort provides a powerful tool for mapping biochemical changes in AD, and a unique opportunity to learn about the association between circulating blood metabolites and brain amyloid-{beta} deposition in AD. OBJECTIVESWe examined 140 serum metabolites and their associations with brain amyloid-{beta} deposition, cognition, and conversion from mild cognitive impairment (MCI) to AD. FINDINGSSerum-based targeted metabolite levels were measured in 1,531 ADNI participants. We performed association analysis of metabolites with brain amyloid-{beta} deposition measured from [18F] Florbetapir PET scans. We identified nine metabolites as significantly associated with amyloid-{beta} deposition after FDR-based multiple comparison correction. Higher levels of one acylcarnitine (C3; propionylcarnitine) and one biogenic amine (kynurenine) were associated with decreased amyloid-{beta} accumulation. However, higher levels of seven phosphatidylcholines (PC) were associated with increased amyloid deposition. In addition, PC ae C44:4 was significantly associated with cognition and conversion from MCI to AD dementia. CONCLUSIONPerturbations in PC and acylcarnitine metabolism may play a role in features intrinsic to AD including amyloid-{beta} deposition and cognitive performance.

neuroscience↗

bioRxiv2020

Purifying selection on noncoding deletions of human regulatory elements detected using their cellular pleiotropy

Genomic deletions provide a powerful loss-of-function model in non-coding regions to assess the role of purifying selection on human noncoding genetic variation. Regulatory element function is char-acterized by non-uniform tissue/cell-type activity, necessarily linking the study of fitness consequences from regulatory variants to their corresponding cellular activity. We used deletions from the 1000 Genomes Project (1000GP) and a callset we generated from genomes of participants in the Alzheimers Disease Neuroimaging Initiative (ADNI) in order to examine whether purifying selection preserves noncoding sites of chromatin accessibility (DHS), histone modification (enhancer, transcribed, polycomb-repressed, heterochromatin), and topologically associated domain loops (TAD-loops). To examine this in a cellular activity-aware manner, we developed a statistical method, Pleiotropy Ratio Score (PlyRS), which calculates a correlation-adjusted count of "cellular pleiotropy" for each noncoding base-pair by analyzing shared regulatory annotations across tissues/cell-types. Comparing real deletion PlyRS values to simulations in a length-matched framework and using genomic covariates in analyses, we found that purifying selection acts to preserve both DHS and enhancer sites, as evident by both depletion of deletions overlapping these annotations and a shift in the allele frequency spectrum of overlapping deletions towards rare alleles. However, we did not find evidence of purifying selection for transcribed, polycomb-repressed, or heterochromatin sites. Additionally, we found evidence that purifying selection is acting on TAD-loop boundary integrity by preserving co-localized CTCF binding sites. Notably, at regions of DHS, enhancer, and CTCF within TAD-loop boundaries we found evidence that both sites of tissue/cell-type-specific activity and sites of cellularly pleiotropic activity are preserved by selection. Significance StatementWe used natural genomic deletions as a loss-of-function model to assess the role of purifying selection in preserving human noncoding regulatory sites. We examined this in a cellular activity-aware manner through development of a statistical method, Pleiotropy Ratio Score (PlyRS), which calculates an adjusted count of "cellular pleiotropy" for each noncoding basepair by analyzing correlations from shared regulatory annotations across tissues/cell-types. By comparing real deletion PlyRS values to simulations, we found that purifying selection acts to preserve both DHS and enhancer sites and TAD-loop boundary integrity by preserving co-localized CTCF binding sites. Notably, we found evidence at these regulatory regions that both sites of tissue/cell-type-specific activity and sites of cellularly pleiotropic activity are preserved by selection.

evolutionary biology↗

bioRxiv2020

Normative modelling using deep autoencoders: a multi-cohort study on mild cognitive impairment and Alzheimer's disease

Normative modelling is an emerging method for quantifying how individuals deviate from the healthy populational pattern. Several machine learning models have been implemented to develop normative models to investigate brain disorders, including regression, support vector machines and Gaussian process models. With the advance of deep learning technology, the use of deep neural networks has also been proposed. In this study, we assessed normative models based on deep autoencoders using structural neuroimaging data from patients with Alzheimers disease (n=206) and mild cognitive impairment (n=354). We first trained the autoencoder on an independent dataset (UK Biobank dataset) with 11,034 healthy controls. Then, we estimated how each patient deviated from this norm and established which brain regions were associated to this deviation. Finally, we compared the performance of our normative model against traditional classifiers. As expected, we found that patients exhibited deviations according to the severity of their clinical condition. The model identified medial temporal regions, including the hippocampus, and the ventricular system as critical regions for the calculation of the deviation score. Overall, the normative model had comparable cross-cohort generalizability to traditional classifiers. In order to promote open science, we are making all scripts and the trained models available to the wider research community.

neuroscience↗

bioRxiv2019

Detecting prodromal Alzheimer’s disease with MRI through deep learning

Deep learning applied to MRI for Alzheimers classification is hypothesized to improve if the deep learning model implicates diseases pathophysiology. The challenge in testing this hypothesis is that large-scale data are required to train this type of model. Here, we overcome this challenge by using a novel data augmentation strategy and show that our MRI-based deep learning model classifies Alzheimers dementia with high accuracy. Moreover, a class activation map was found dominated by signal from the hippocampal formation, a site where Alzheimers pathophysiology begins. Next, we tested the models performance in prodromal Alzheimers when patients present with mild cognitive impairment (MCI). We retroactively dichotomized a large cohort of MCI patients who were followed for up to 10 years into those with and without prodromal Alzheimers at baseline and used the dementia-derived model to generate individual deep learning MRI scores. We compared the two groups on these scores, and on other biomarkers of amyloid pathology, tau pathology, and neurodegeneration. The deep learning MRI scores outperformed nearly all other biomarkers, including--unexpectedly--biomarkers of amyloid or tau pathology, in classifying prodromal disease and in predicting clinical progression. Providing a mechanistic explanation, the deep learning MRI scores were found to be linked to regional tau pathology, through investigations using cross-sectional, longitudinal, premortem and postmortem data. Our findings validate that a diseases known pathophysiology can improve the design and performance of deep learning models. Moreover, by showing that deep learning can extract useful biomarker information from conventional MRIs, the advantages of this model extend practically, potentially reducing patient burden, risk, and cost.

neuroscience↗