scholarly journals Deep Generative Analysis for Task-Based Functional MRI Experiments

2021 ◽  
Author(s):  
Daniela F de Albuquerque ◽  
Jack Goffinet ◽  
Rachael Wright ◽  
John M Pearson
Keyword(s):  
Additive Model ◽  
Brain Regions ◽  
Generative Models ◽  
Statistical Testing ◽  
Single Subject ◽  
Great Success ◽  
Current Standard ◽  
Brain Volumes ◽  
Fmri Analysis ◽  
Post Hoc

While functional magnetic resonance imaging (fMRI) remains one of the most widespread and important methods in basic and clinical neuroscience, the data it produces---time series of brain volumes---continue to pose daunting analysis challenges. The current standard ("mass univariate") approach involves constructing a matrix of task regressors, fitting a separate general linear model at each volume pixel ("voxel"), computing test statistics for each model, and correcting for false positives post hoc using bootstrap or other resampling methods. Despite its simplicity, this approach has enjoyed great success over the last two decades due to: 1) its ability to produce effect maps highlighting brain regions whose activity significantly correlates with a given variable of interest; and 2) its modeling of experimental effects as separable and thus easily interpretable. However, this approach suffers from several well-known drawbacks, namely: inaccurate assumptions of linearity and noise Gaussianity; a limited ability to capture individual effects and variability; and difficulties in performing proper statistical testing secondary to independently fitting voxels. In this work, we adopt a different approach, modeling entire volumes directly in a manner that increases model flexibility while preserving interpretability. Specifically, we use a generalized additive model (GAM) in which the effects of each regressor remain separable, the product of a spatial map produced by a variational autoencoder and a (potentially nonlinear) gain modeled by a covariate-specific Gaussian Process. The result is a model that yields group-level effect maps comparable or superior to the ones obtained with standard fMRI analysis software while also producing single-subject effect maps capturing individual differences. This suggests that generative models with a decomposable structure might offer a more flexible alternative for the analysis of task-based fMRI data.

scholarly journals Automated quantitative MRI volumetry reports support diagnostic interpretation in dementia: a multi-rater, clinical accuracy study

2021 ◽  
Author(s):  
Hugh G. Pemberton ◽  
◽  
Olivia Goodkin ◽  
Ferran Prados ◽  
Ravi K. Das ◽  
...  
Keyword(s):  
Visual Assessment ◽  
Quantitative Mri ◽  
Interrater Agreement ◽  
Clinical Image ◽  
Single Subject ◽  
Volume Loss ◽  
Brain Volumes ◽  
Additional Information ◽  
Clinical Accuracy ◽  
Accuracy Study

Abstract Objectives We examined whether providing a quantitative report (QReport) of regional brain volumes improves radiologists’ accuracy and confidence in detecting volume loss, and in differentiating Alzheimer’s disease (AD) and frontotemporal dementia (FTD), compared with visual assessment alone. Methods Our forced-choice multi-rater clinical accuracy study used MRI from 16 AD patients, 14 FTD patients, and 15 healthy controls; age range 52–81. Our QReport was presented to raters with regional grey matter volumes plotted as percentiles against data from a normative population (n = 461). Nine raters with varying radiological experience (3 each: consultants, registrars, ‘non-clinical image analysts’) assessed each case twice (with and without the QReport). Raters were blinded to clinical and demographic information; they classified scans as ‘normal’ or ‘abnormal’ and if ‘abnormal’ as ‘AD’ or ‘FTD’. Results The QReport improved sensitivity for detecting volume loss and AD across all raters combined (p = 0.015* and p = 0.002*, respectively). Only the consultant group’s accuracy increased significantly when using the QReport (p = 0.02*). Overall, raters’ agreement (Cohen’s κ) with the ‘gold standard’ was not significantly affected by the QReport; only the consultant group improved significantly (κs 0.41➔0.55, p = 0.04*). Cronbach’s alpha for interrater agreement improved from 0.886 to 0.925, corresponding to an improvement from ‘good’ to ‘excellent’. Conclusion Our QReport referencing single-subject results to normative data alongside visual assessment improved sensitivity, accuracy, and interrater agreement for detecting volume loss. The QReport was most effective in the consultants, suggesting that experience is needed to fully benefit from the additional information provided by quantitative analyses. Key Points • The use of quantitative report alongside routine visual MRI assessment improves sensitivity and accuracy for detecting volume loss and AD vs visual assessment alone. • Consultant neuroradiologists’ assessment accuracy and agreement (kappa scores) significantly improved with the use of quantitative atrophy reports. • First multi-rater radiological clinical evaluation of visual quantitative MRI atrophy report for use as a diagnostic aid in dementia.

scholarly journals Conditional out-of-distribution generation for unpaired data using transfer VAE

2020 ◽  
Vol 36 (Supplement_2) ◽  
pp. i610-i617
Author(s):  
Mohammad Lotfollahi ◽  
Mohsen Naghipourfar ◽  
Fabian J Theis ◽  
F Alexander Wolf
Keyword(s):  
Single Cell ◽  
Generative Models ◽  
Response To Treatment ◽  
High Dimensional ◽  
Compact Representation ◽  
Hair Color ◽  
Great Success ◽  
Cell Type ◽  
Style Transfer ◽  
Cell Type Specific

Abstract Motivation While generative models have shown great success in sampling high-dimensional samples conditional on low-dimensional descriptors (stroke thickness in MNIST, hair color in CelebA, speaker identity in WaveNet), their generation out-of-distribution poses fundamental problems due to the difficulty of learning compact joint distribution across conditions. The canonical example of the conditional variational autoencoder (CVAE), for instance, does not explicitly relate conditions during training and, hence, has no explicit incentive of learning such a compact representation. Results We overcome the limitation of the CVAE by matching distributions across conditions using maximum mean discrepancy in the decoder layer that follows the bottleneck. This introduces a strong regularization both for reconstructing samples within the same condition and for transforming samples across conditions, resulting in much improved generalization. As this amount to solving a style-transfer problem, we refer to the model as transfer VAE (trVAE). Benchmarking trVAE on high-dimensional image and single-cell RNA-seq, we demonstrate higher robustness and higher accuracy than existing approaches. We also show qualitatively improved predictions by tackling previously problematic minority classes and multiple conditions in the context of cellular perturbation response to treatment and disease based on high-dimensional single-cell gene expression data. For generic tasks, we improve Pearson correlations of high-dimensional estimated means and variances with their ground truths from 0.89 to 0.97 and 0.75 to 0.87, respectively. We further demonstrate that trVAE learns cell-type-specific responses after perturbation and improves the prediction of most cell-type-specific genes by 65%. Availability and implementation The trVAE implementation is available via github.com/theislab/trvae. The results of this article can be reproduced via github.com/theislab/trvae_reproducibility.

scholarly journals Reported Hearing Loss in Alzheimer’s Disease Is Associated With Loss of Brainstem and Cerebellar Volume

2021 ◽  
Vol 15 ◽  
Author(s):  
Daniel A. Llano ◽  
Susanna S. Kwok ◽  
Viswanath Devanarayan ◽  
Keyword(s):  
Alzheimer’S Disease ◽  
Alzheimer's Disease ◽  
Hearing Loss ◽  
Normal Subjects ◽  
Neural Substrates ◽  
Epidemiological Studies ◽  
Brain Regions ◽  
Pathological State ◽  
Brain Volumes ◽  
The Relationship

Multiple epidemiological studies have revealed an association between presbycusis and Alzheimer’s Disease (AD). Unfortunately, the neurobiological underpinnings of this relationship are not clear. It is possible that the two disorders share a common, as yet unidentified, risk factor, or that hearing loss may independently accelerate AD pathology. Here, we examined the relationship between reported hearing loss and brain volumes in normal, mild cognitive impairment (MCI) and AD subjects using a publicly available database. We found that among subjects with AD, individuals that reported hearing loss had smaller brainstem and cerebellar volumes in both hemispheres than individuals without hearing loss. In addition, we found that these brain volumes diminish in size more rapidly among normal subjects with reported hearing loss and that there was a significant interaction between cognitive diagnosis and the relationship between reported hearing loss and these brain volumes. These data suggest that hearing loss is linked to brainstem and cerebellar pathology, but only in the context of the pathological state of AD. We hypothesize that the presence of AD-related pathology in both the brainstem and cerebellum creates vulnerabilities in these brain regions to auditory deafferentation-related atrophy. These data have implications for our understanding of the potential neural substrates for interactions between hearing loss and AD.

scholarly journals Functional MRI and the Study of Human Consciousness

2002 ◽  
Vol 14 (6) ◽  
pp. 818-831 ◽  
Author(s):  
Dan Lloyd
Keyword(s):  
Spatial Working Memory ◽  
Cognitive Task ◽  
Response Competition ◽  
Temporal Cortex ◽  
Brain Regions ◽  
Functional Brain Imaging ◽  
Future Research ◽  
Single Subject ◽  
Neural Activation ◽  
Human Consciousness

Functional brain imaging offers new opportunities for the study of that most pervasive of cognitive conditions, human consciousness. Since consciousness is attendant to so much of human cognitive life, its study requires secondary analysis of multiple experimental datasets. Here, four preprocessed datasets from the National fMRI Data Center are considered: Hazeltine et al., Neural activation during response competition; Ishai et al., The representation of objects in the human occipital and temporal cortex; Mechelli et al., The effects of presentation rate during word and pseudoword reading; and Postle et al., Activity in human frontal cortex associated with spatial working memory and saccadic behavior. The study of consciousness also draws from multiple disciplines. In this article, the philosophical subdiscipline of phenomenology provides initial characterization of phenomenal structures conceptually necessary for an analysis of consciousness. These structures include phenomenal intentionality, phenomenal superposition, and experienced temporality. The empirical predictions arising from these structures require new interpretive methods for their confirmation. These methods begin with single-subject (preprocessed) scan series, and consider the patterns of all voxels as potential multivariate encodings of phenomenal information. Twenty-seven subjects from the four studies were analyzed with multivariate methods, revealing analogues of phenomenal structures, particularly the structures of temporality. In a second interpretive approach, artificial neural networks were used to detect a more explicit prediction from phenomenology, namely, that present experience contains and is inflected by past states of awareness and anticipated events. In all of 21 subjects in this analysis, nets were successfully trained to extract aspects of relative past and future brain states, in comparison with statistically similar controls. This exploratory study thus concludes that the proposed methods for “neurophenomenology” warrant further application, including the exploration of individual differences, multivariate differences between cognitive task conditions, and exploration of specific brain regions possibly contributing to the observations. All of these attractive questions, however, must be reserved for future research.

scholarly journals Relationship Between Verbal Fluency Performance, Fight Exposure, and Subcortical Region Brain Volumes in Professional Fighters

2019 ◽  
Vol 34 (5) ◽  
pp. 735-735
Author(s):  
L Bennett ◽  
C Bernick ◽  
S Banks
Keyword(s):  
Verbal Fluency ◽  
Brain Structures ◽  
Brain Regions ◽  
Health Study ◽  
Semantic Fluency ◽  
Cognitive Changes ◽  
Brain Volumes ◽  
Subcortical Brain ◽  
Subcortical Region ◽  
Subcortical Brain Structures

Abstract Purpose Verbal fluency performance has been shown to be sensitive to preclinical cognitive changes in neurodegenerative diseases and may detect early, trauma-related cognitive and volumetric changes amongst professional fighters. Baseline verbal fluency performance and volumes of relevant subcortical brain structures were expected to decline as number of professional fights (NoPF) increased, while controlling for education. Methods Baseline letter and semantic fluency performance, NoPF, and structural brain imaging from 548 active and retired fighters who participated in the Professional Fighters Brain Health Study were considered. ANCOVAs were conducted to assess differences in verbal fluency performance by NOPF, while controlling for years of education. Number of professional fights were stratified into low (0-20 fights), medium (21-40 fights), and high (41 or more fights). Results Semantic fluency performance differed across the three levels of NoPF (F(2, 542)=4.56; p<.02). In addition, significant positive correlations between semantic fluency performance and volumes in the following regions were observed: left thalamus, left putamen, left pallidum, bilateral caudates, bilateral amygdalae, bilateral hippocampi, and bilateral accumbens (all p’s<.05). In contrast, letter fluency performance was not significantly associated with NoPF or volumes of relevant subcortical brain structures (all p’s>.05). Conclusion Semantic fluency may be low-cost, easy-to-administer harbinger of emerging cognitive dysfunction and lower volumes in related subcortical brain regions. Additional assessment of clinical utility is necessary.

scholarly journals Brain structural correlates of functional capacity in first-episode psychosis

2020 ◽  
Vol 10 (1) ◽  
Author(s):  
Erkan Alkan ◽  
Geoff Davies ◽  
Kathy Greenwood ◽  
Simon L. Evans
Keyword(s):  
Functional Capacity ◽  
Gray Matter Volume ◽  
Brain Regions ◽  
First Episode Psychosis ◽  
First Episode ◽  
Control Group ◽  
Neural Basis ◽  
Antipsychotic Therapy ◽  
Brain Volumes ◽  
Episode Psychosis

Abstract Impaired functional capacity is a core feature of schizophrenia and presents even in first-episode psychosis (FEP) patients. Impairments in daily functioning tend to persist despite antipsychotic therapy but their neural basis is less clear. Previous studies suggest that volume loss in frontal cortex might be an important contributor, but findings are inconsistent. We aimed to comprehensively investigate the brain structural correlates of functional capacity in FEP using MRI and a reliable objective measure of functioning [University of California, San Diego Performance-Based Skills Assessment (UPSA)]. In a sample of FEP (n = 39) and a well-matched control group (n = 21), we measured cortical thickness, gray matter volume, and white matter tract integrity (fractional anisotropy, FA) within brain regions implicated by previous work. The FEP group had thinner cortex in various frontal regions and fusiform, and reduced FA in inferior longitudinal fasciculus (ILF). In FEP, poorer functional capacity correlated with reduced superior frontal volume and lower FA in left ILF. Importantly, frontal brain volumes and integrity of the ILF were identified as the structural correlates of functional capacity in FEP, controlling for other relevant factors. These findings enhance mechanistic understanding of functional capacity deficits in schizophrenia by specifying the underlying neural correlates. In future, this could help inform intervention strategies.

scholarly journals Review of Twin and Family Studies on Neuroanatomic Phenotypes and Typical Neurodevelopment

2007 ◽  
Vol 10 (5) ◽  
pp. 683-694 ◽  
Author(s):  
J. Eric Schmitt ◽  
Lisa T. Eyler ◽  
Jay N. Giedd ◽  
William S. Kremen ◽  
Kenneth S. Kendler ◽  
...  
Keyword(s):  
Brain Structure ◽  
Emotional Regulation ◽  
Behavioral Genetics ◽  
Genetic Relationships ◽  
Twin Studies ◽  
Brain Regions ◽  
Phenotypic Variance ◽  
Imaging Data ◽  
Brain Volumes ◽  
Large Brain

AbstractThis article reviews the extant twin studies employing magnetic resonance imaging data (MRI), with an emphasis on studies of populationbased samples. There have been approximately 75 twin reports using MRI, with somewhat under half focusing on typical brain structure. Of these, most are samples of adults. For large brain regions such as lobar volumes, the heritabilities of large brain volumes are consistently high, with genetic factors accounting for at least half of the phenotypic variance. The role of genetics in generating individual differences in the volumes of small brain regions is less clear, mostly due to a dearth of information, but rarely because of disagreement between studies. Multivariate analyses show strong genetic relationships between brain regions. Cortical regions involved in language, executive function, and emotional regulation appear to be more heritable than other areas. Studies of brain shape also show significant, albeit lower, genetic effects on population variance. Finally, there is evidence of significant genetically mediated relationships between intelligence and brain structure. At present, the majority of twin imaging studies are limited by sample sizes small by the standards of behavioral genetics; nevertheless the literature at present represents a pioneering effort in the pursuit of answers to many challenging neurobiological questions.

scholarly journals Inferring the Genetic Influences on Psychological Traits Using MRI Connectivity Predictive Models: Demonstration with Cognition

2019 ◽  
Author(s):  
Alexander S. Hatoum ◽  
Andrew E. Reineberg ◽  
Philip A. Kragel ◽  
Tor D. Wager ◽  
Naomi P. Friedman
Keyword(s):  
Predictive Models ◽  
Model Development ◽  
Genetic Correlations ◽  
Additive Model ◽  
Brain Regions ◽  
Imaging Genetics ◽  
Psychological Traits ◽  
Genetics Research ◽  
Brain Models ◽  
And Behavior

AbstractGenetic correlations between brain and behavioral phenotypes in analyses from major genetic consortia have been weak and mostly non-significant. fMRI models of systems-level brain patterns may help improve our ability to link genes, brains, and behavior by identifying reliable and reproducible endophenotypes. Work using connectivity-based predictive modeling (CBPM) has generated brain-based proxies of behavioral and neuropsychological variables. If such models capture activity in inherited brain systems, they may offer a more powerful link between genes and behavior. As a proof of concept, we develop models predicting intelligence (IQ) based on fMRI connectivity and test their effectiveness as endophenotypes. We link brain and IQ in a model development dataset of N=3,000 individuals; and test the genetic correlations between brain models and measured IQ in a genetic validation sample of N=13,092 individuals from the UKBiobank. We compare an additive connectivity-based model to multivariate LASSO and ridge models phenotypically and genetically. We also compare these approaches to single “candidate” brain areas. We find that predictive brain models were significantly phenotypically correlated with IQ and showed much stronger correlations than individual edges. Further, brain models were more heritable than single brain regions (h2=.155-.181) and capture about half of the genetic variance in IQ (rG=.422-.576), while rGs with single brain measures were smaller and non-significant. For the different approaches, LASSO and Ridge were similarly predictive, with slightly weaker performance of the additive model. LASSO model weights were highly theoretically interpretable and replicated known brain IQ associations. Finally, functional connectivity models trained in midlife showed genetic correlations with early life correlates of IQ, suggesting some stability in the prediction of fMRI models. We conclude that multi-system predictive models hold promise as imaging endophenotypes that offer complex and theoretically relevant conclusions for future imaging genetics research.

Sign in / Sign up

Export Citation Format

Share Document