Human Genetics Publications

Testing Spatial Correlation Between Brain Maps Using Dual Permutation

Document Type

Article

Publication Date

10-5-2026

Abstract

Population mean maps (PMMs) are commonly shared in neuroimaging, allowing researchers to assess relationships between studies by computing spatial correlations between maps. However, determining the statistical significance of these correlations is challenging because spatial autocorrelation and voxel-wise noise variance are often unknown and spatially heterogeneous. Here we introduce Dual Permutation (DP), a flexible non-parametric method for computing p-values for spatial correlations between PMMs derived from distinct subject cohorts. DP operates by permuting or sign-flipping subjects within each modality to create surrogates that preserve the spatial structure of the noise. The resulting surrogate PMMs are used to generate an empirical null distribution of the correlation coefficient. DP builds upon unimodal subject-level permutation, a well-established approach in neuroimaging that preserves spatially heterogeneous noise structure. We show that the problem is more complex than generally appreciated because different null distributions can arise when signal is present in one modality but not the other; we introduce a method called Maximum Difference in Variance (MaxDV) to address this issue. We derive analytical results demonstrating that DP preserves the variance of the correlation coefficient under the null hypothesis across a broad range of spatial autocorrelation, voxel scaling, noise characteristics, and group-level designs. Using a diverse set of real neuroimaging datasets spanning volumetric, surface-based, and region-of-interest analyses, we show that DP accurately controls the false positive rate. Although DP requires subject-level data and cannot be applied when only PMMs are available, we also demonstrate that PMM-only analyses can produce both false positive and false negative findings. A publicly available implementation of DP is available in FreeSurfer as the mri_dualperm tool.

Comments

© 2026 The Authors. Published under a Creative Commons Attribution 4.0 International (CC BY 4.0) license.

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Publication Title

Imaging Neuroscience

DOI

10.1162/IMAG.a.1418

Link to Full Text

Share

COinS