Dataset Information

Faster permutation inference in brain imaging.

ABSTRACT: Permutation tests are increasingly being used as a reliable method for inference in neuroimaging analysis. However, they are computationally intensive. For small, non-imaging datasets, recomputing a model thousands of times is seldom a problem, but for large, complex models this can be prohibitively slow, even with the availability of inexpensive computing power. Here we exploit properties of statistics used with the general linear model (GLM) and their distributions to obtain accelerations irrespective of generic software or hardware improvements. We compare the following approaches: (i) performing a small number of permutations; (ii) estimating the p-value as a parameter of a negative binomial distribution; (iii) fitting a generalised Pareto distribution to the tail of the permutation distribution; (iv) computing p-values based on the expected moments of the permutation distribution, approximated from a gamma distribution; (v) direct fitting of a gamma distribution to the empirical permutation distribution; and (vi) permuting a reduced number of voxels, with completion of the remainder using low rank matrix theory. Using synthetic data we assessed the different methods in terms of their error rates, power, agreement with a reference result, and the risk of taking a different decision regarding the rejection of the null hypotheses (known as the resampling risk). We also conducted a re-analysis of a voxel-based morphometry study as a real-data example. All methods yielded exact error rates. Likewise, power was similar across methods. Resampling risk was higher for methods (i), (iii) and (v). For comparable resampling risks, the method in which no permutations are done (iv) was the absolute fastest. All methods produced visually similar maps for the real data, with stronger effects being detected in the family-wise error rate corrected maps by (iii) and (v), and generally similar to the results seen in the reference set. Overall, for uncorrected p-values, method (iv) was found the best as long as symmetric errors can be assumed. In all other settings, including for familywise error corrected p-values, we recommend the tail approximation (iii). The methods considered are freely available in the tool PALM - Permutation Analysis of Linear Models.

SUBMITTER: Winkler AM

PROVIDER: S-EPMC5035139 | biostudies-literature | 2016 Nov

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Faster permutation inference in brain imaging.

Winkler Anderson M AM Ridgway Gerard R GR Douaud Gwenaëlle G Nichols Thomas E TE Smith Stephen M SM

NeuroImage 20160607

Permutation tests are increasingly being used as a reliable method for inference in neuroimaging analysis. However, they are computationally intensive. For small, non-imaging datasets, recomputing a model thousands of times is seldom a problem, but for large, complex models this can be prohibitively slow, even with the availability of inexpensive computing power. Here we exploit properties of statistics used with the general linear model (GLM) and their distributions to obtain accelerations irre ...[more]

PMID: 27288322

Similar Datasets

Project description:There is considerable scientific interest in understanding how cell assemblies-the long-presumed computational motif-are organized so that the brain can generate intelligent cognition and flexible behavior. The Theory of Connectivity proposes that the origin of intelligence is rooted in a power-of-two-based permutation logic (N = 2 i -1), producing specific-to-general cell-assembly architecture capable of generating specific perceptions and memories, as well as generalized knowledge and flexible actions. We show that this power-of-two-based permutation logic is widely used in cortical and subcortical circuits across animal species and is conserved for the processing of a variety of cognitive modalities including appetitive, emotional and social information. However, modulatory neurons, such as dopaminergic (DA) neurons, use a simpler logic despite their distinct subtypes. Interestingly, this specific-to-general permutation logic remained largely intact although NMDA receptors-the synaptic switch for learning and memory-were deleted throughout adulthood, suggesting that the logic is developmentally pre-configured. Moreover, this computational logic is implemented in the cortex via combining a random-connectivity strategy in superficial layers 2/3 with nonrandom organizations in deep layers 5/6. This randomness of layers 2/3 cliques-which preferentially encode specific and low-combinatorial features and project inter-cortically-is ideal for maximizing cross-modality novel pattern-extraction, pattern-discrimination and pattern-categorization using sparse code, consequently explaining why it requires hippocampal offline-consolidation. In contrast, the nonrandomness in layers 5/6-which consists of few specific cliques but a higher portion of more general cliques projecting mostly to subcortical systems-is ideal for feedback-control of motivation, emotion, consciousness and behaviors. These observations suggest that the brain's basic computational algorithm is indeed organized by the power-of-two-based permutation logic. This simple mathematical logic can account for brain computation across the entire evolutionary spectrum, ranging from the simplest neural networks to the most complex.

Project description:BackgroundOptimization of mixture analyses is critical to assess potential impacts of human exposures to multiple pollutants. Single-index regression methods quantify total mixture association and chemical component contributions. Single-index methods include several variants of quantile g-computation (QGC) and weighted quantile sum regression (WQSr), though each has limitations.ObjectivesWe developed a novel permutation test for WQSr and compared its performance to extant versions of WQSr and QGC in simulation studies and an analysis of prenatal phthalates and childhood cognition.MethodsWQSr uses ensemble nonlinear optimization to identify weights for mixture exposures in an index associated with the outcome in a prespecified direction, with ensembles based on bootstrap resampling (WQSBS) or random subsetting of exposures (WQSRS). Statistical significance can be assessed without splitting the data (Nosplit), by splitting the data into training and test sets (Split), by repeatedly holding out test sets (RH), or by using a novel permutation test (PT) to obtain a more accurate p-value. QGC instead provides a sum mixture coefficient and component coefficients with no constraints on direction. In simulations, we compared false positive rates (FPR) and power to detect true associations and accuracy in estimating mixture weights. We also estimated associations between prenatal phthalate mixtures and childhood IQ in the Conditions Affecting Neurocognitive Development and Learning in Early Childhood cohort using each method.ResultsFPR was well controlled at ≤7% in all but the Nosplit WQSr variants. Among these methods, the WQSBS and WQSRS PT variants had the highest power (89%-98%), with lower power for QGC (85%-93%) and RH (60%-97%) or Split WQSr variants (40%-78%). WQSr methods estimated mixture weights 2-4 times more accurately than the QGC method. Coefficients for mixture associations with full scale IQ varied 3- to 4-fold across analytic methods.DiscussionWQSr paired with our novel permutation test best balanced power and false positive rate when assessing a mixture effect. As new methods develop, each should be examined for performance and applicability. https://doi.org/10.1289/EHP10570.

Project description:Our knowledge of the anatomical organization of the human brain in health and disease draws heavily on the study of patients with focal brain lesions. Historically the first method of mapping brain function, it is still potentially the most powerful, establishing the necessity of any putative neural substrate for a given function or deficit. Great inferential power, however, carries a crucial vulnerability: without stronger alternatives any consistent error cannot be easily detected. A hitherto unexamined source of such error is the structure of the high-dimensional distribution of patterns of focal damage, especially in ischaemic injury-the commonest aetiology in lesion-deficit studies-where the anatomy is naturally shaped by the architecture of the vascular tree. This distribution is so complex that analysis of lesion data sets of conventional size cannot illuminate its structure, leaving us in the dark about the presence or absence of such error. To examine this crucial question we assembled the largest known set of focal brain lesions (n = 581), derived from unselected patients with acute ischaemic injury (mean age = 62.3 years, standard deviation = 17.8, male:female ratio = 0.547), visualized with diffusion-weighted magnetic resonance imaging, and processed with validated automated lesion segmentation routines. High-dimensional analysis of this data revealed a hidden bias within the multivariate patterns of damage that will consistently distort lesion-deficit maps, displacing inferred critical regions from their true locations, in a manner opaque to replication. Quantifying the size of this mislocalization demonstrates that past lesion-deficit relationships estimated with conventional inferential methodology are likely to be significantly displaced, by a magnitude dependent on the unknown underlying lesion-deficit relationship itself. Past studies therefore cannot be retrospectively corrected, except by new knowledge that would render them redundant. Positively, we show that novel machine learning techniques employing high-dimensional inference can nonetheless accurately converge on the true locus. We conclude that current inferences about human brain function and deficits based on lesion mapping must be re-evaluated with methodology that adequately captures the high-dimensional structure of lesion data.

Dataset Information

Faster permutation inference in brain imaging.

Publications

Faster permutation inference in brain imaging.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets