A simple permutation‐based test of intermodal correspondence

Abstract

Many key findings in neuroimaging studies involve similarities between brain maps, but statistical methods used to measure these findings have varied. Current state‐of‐the‐art methods involve comparing observed group‐level brain maps (after averaging intensities at each image location across multiple subjects) against spatial null models of these group‐level maps. However, these methods typically make strong and potentially unrealistic statistical assumptions, such as covariance stationarity. To address these issues, in this article we propose using subject‐level data and a classical permutation testing framework to test and assess similarities between brain maps. Our method is comparable to traditional permutation tests in that it involves randomly permuting subjects to generate a null distribution of intermodal correspondence statistics, which we compare to an observed statistic to estimate a p‐value. We apply and compare our method in simulated and real neuroimaging data from the Philadelphia Neurodevelopmental Cohort. We show that our method performs well for detecting relationships between modalities known to be strongly related (cortical thickness and sulcal depth), and it is conservative when an association would not be expected (cortical thickness and activation on the n‐back working memory task). Notably, our method is the most flexible and reliable for localizing intermodal relationships within subregions of the brain and allows for generalizable statistical inference.

Spatially-enhanced clusterwise inference for testing and localizing intermodal correspondence

“…To our knowledge, among the existing methods for testing intermodal correspondence, Burt et al [7]’s BrainSMASH is the only widely used method that includes a model of the spatial autocorrelation structure (although Alexander-Bloch et al [6], Weinstein et al [9] handle this in other ways that do not require explicit spatial modeling). Burt et al’s null hypothesis for testing similarity between group-level (i.e.…”

“…between structural and functional brain measurements). Methodological developments in this area have included statistical estimation of individual-level maps measuring intermodal coupling [1, 2, 3, 4, 5] and hypothesis tests for establishing statistical significance of intermodal associations [6, 7, 8, 9].…”

“…In our current work, we propose a novel hypothesis testing method that involves a combination of approaches from Weinstein et al [9] (SPICE) and recent work by Park and Fiecas [10] that explicitly models spatial autocorrelation in cluster-level inference. Using Park and Fiecas [10]’s spatial modeling framework, our method leverages spatial information through cluster-level inference after adjusting for differences in the spatial autocorrelation structure of each modality.…”

“…Using Park and Fiecas [10]’s spatial modeling framework, our method leverages spatial information through cluster-level inference after adjusting for differences in the spatial autocorrelation structure of each modality. Using a permutation framework similar to our previous work [9], our new method controls type I error rates at the nominal level and exhibits high statistical power for both global testing and localization of intermodal correspondence, as assessed in simulations and analyses of structure-function coupling in data from the Philadelphia Neurodevelopmental Cohort.…”

Similarity-Based Multimodal Regression

“…Our second PNC application uses a subset of 912 PNC subjects with high-quality cortical thickness (CT) and sulcal depth (SD) measurements computed from T1-weighted images (demographic details in Table 2 under “PNC Cortical Structure”). The acquisition and preprocessing for these data was originally described in Vandekar et al (2015) and in subsequent studies (Vandekar et al, 2016; Weinstein et al, 2021). Cortical reconstruction of the T1-weighted structural images was completed using FreeSurfer (version 5.3).…”