Spatially-enhanced clusterwise inference for testing and localizing intermodal correspondence

Abstract

With the increasing availability of neuroimaging data from multiple modalities—each providing a different lens through which to study brain structure or function—new techniques for comparing, integrating, and interpreting information within and across modalities have emerged. Recent developments include hypothesis tests of associations between neuroimaging modalities, which can be used to determine the statistical significance of intermodal associations either throughout the entire brain or within anatomical subregions or functional networks. While these methods provide a crucial foundation for inference on intermodal relationships, they cannot be used to answer questions about where in the brain these associations are most pronounced. In this paper, we introduce a new method, called CLEAN-R, that can be used both to test intermodal correspondence throughout the brain and also to localize this correspondence. Our method involves first adjusting for the underlying spatial autocorrelation structure within each modality before aggregating information within small clusters to construct a map of spatially enhanced test statistics. Using structural and functional magnetic resonance imaging data from a subsample of children and adolescents from the Philadelphia Neurodevelopmental Cohort, we conduct simulations and data analyses where we illustrate the high statistical power and nominal type I error levels of our method. By constructing an interpretable map of group-level correspondence using spatially-enhanced test statistics, our method offers insights beyond those provided by earlier methods.

Spatial-extent inference for testing variance components in reliability and heritability studies

“…Our results showed that it is necessary to use both cluster-wise enhancement and spatial autocorrelation modeling to increase sensitivity in testing variance components, which aligns with previous findings [12,13]. It was supported by comparing CLEAN-V to simpler methods: massive univariate analysis, CLEAN-V without spatial correlation, and CLEAN-V without cluster enhancement.…”

“…cortical thickness) where 'borrowing' information across spatial domains is expected to improve power. Also, a comprehensive exploratory analysis by Weinstein et al [13] reveals that applying exponential spatial autocorrelation function (SACF) is reasonable in many imaging modalities mapped onto the cortical surface. While applying the same assumption to volumetric (3D) imaging data could be questioned, we showed that a simplified version of CLEAN ('CLEAN-V without spatial correlation') also results in a dramatic increase in power and sensitivity compared to massive univariate analysis, which can be adopted easily.…”

“…In this way, we can find the closely optimal cluster size for each vertex, which is useful as true areas of signals are unknown a priori. Following Park and Fiecas [12] and Weinstein et al [13], we consider r 1 = 1mm, r 2 = 2mm, . .…”

“…Motivated by existing efforts in neuroimaging literature [12,13,24], we used task-fMRI data from HCP to design data-driven simulations to evaluate the performance of CLEAN-V compared to the competitors. This is particularly useful in evaluating the robustness of the method under potential model misspecifications because CLEAN-V assumes stationarity and isotropy of underlying spatial autocorrelation, which may not hold.…”

“…More recently, the CLEAN method proposed by Park and Fiecas [12] showed that additional gain in sensitivity is achieved in GLMs by cluster-enhanced statistic, which is constructed from multivariate test statistics by applying spatial Gaussian process to the residual noise. This idea has been extended to CLEAN-R [13] in test-ing and localizing correspondence between two imaging modalities by modeling modality-specific spatial autocorrelation and constructing cluster-enhanced test statistics for correlation parameters.…”