Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 Sep:178:540-551.
doi: 10.1016/j.neuroimage.2018.05.070. Epub 2018 Jun 1.

On testing for spatial correspondence between maps of human brain structure and function

Aaron F Alexander-Bloch  1 , Haochang Shou  2 , Siyuan Liu  3 , Theodore D Satterthwaite  4 , David C Glahn  5 , Russell T Shinohara  2 , Simon N Vandekar  2 , Armin Raznahan  3
Affiliations

On testing for spatial correspondence between maps of human brain structure and function

Aaron F Alexander-Bloch et al. Neuroimage. 2018 Sep.

Abstract

A critical issue in many neuroimaging studies is the comparison between brain maps. Nonetheless, it remains unclear how one should test hypotheses focused on the overlap or spatial correspondence between two or more brain maps. This "correspondence problem" affects, for example, the interpretation of comparisons between task-based patterns of functional activation, resting-state networks or modules, and neuroanatomical landmarks. To date, this problem has been addressed with remarkable variability in terms of methodological approaches and statistical rigor. In this paper, we address the correspondence problem using a spatial permutation framework to generate null models of overlap by applying random rotations to spherical representations of the cortical surface, an approach for which we also provide a theoretical statistical foundation. We use this method to derive clusters of cognitive functions that are correlated in terms of their functional neuroatomical substrates. In addition, using publicly available data, we formally demonstrate the correspondence between maps of task-based functional activity, resting-state fMRI networks and gyral-based anatomical landmarks. We provide open-access code to implement the methods presented for two commonly-used tools for surface based cortical analysis (https://www.github.com/spin-test). This spatial permutation approach constitutes a useful advance over widely-used methods for the comparison of cortical maps, thereby opening new possibilities for the integration of diverse neuroimaging data.

PubMed Disclaimer

Figures

Figure 1
Figure 1
A schematic of the permutation procedure. A) As an illustration, the Desikan atlas is shown in the original space (top left) and spherical space (top right). Each color corresponds to different regions. The spherical coordinates are rotated (mid right, bottom right) and the projected back onto the anatomical surface (mid left, bottom left). B) The degree of similarity between the original parcellation and the rotated parcellations were estimated using the normalized mutual information (NMI). The probability density distributions of this statistic are shown for 100, 500, and 1000 rotations, as well as lines marking the 95th percentile of each distribution. C) A Q-Q plot of the two independent distributions of 1000 rotations each.
Figure 2
Figure 2
Correlation structure and significant correlations between Neurosynth meta-analytic activation patterns associated with 120 cognitive terms. A) Heat-map shows 120×120 correlation matrix. Terms are organized according to hierarchical clustering, with the resulting dendrogram shown to the top and to the left of the correlation matrix. Colors correspond to correlation coefficient, as shown in color key on top left. The color key also shows the frequency distribution of the correlations that comprise the matrix. Labels of the terms are shown to the right and to the bottom of the matrix, with the odd number labels shown on the bottom and the even number labels shown on the right (the order of the terms is “fear”, “anxiety”, “stress”, “arousal”, “valence,” etc.). B) Network illustration where the significant connections are illustrated as edges between the terms, which are illustrated as nodes. The resulting network is comprised of 8 disconnected components; edges exist within each component’s nodes, but there are no edges between components.
Figure 3
Figure 3
Spatial relationship between regions based on gyral landmarks (Desikan Atlas), intrinsic functional connectivity networks (Yeo Atlas), and task-based fMRI brain maps (Neurosynth meta-analyses). A) Representation of the Desikan Atlas, derived from manually identifying 34 in each hemisphere based on gyral landmarks, using 40 high resolution structural MRI scans. B) Representation of the Yeo atlas, derived by identifying 7 resting-state functional networks using a mixture model of 1000 resting-state fMRI scans. C) The normalized mutation information between the Yeo and Desikan Atlas, a measure of the similarity of the two atlases, for the original data as well as the probability density distribution of 1000 rotational permutations. The P-value is calculated as the frequency with which the permuted NMI equals or exceeds the actual NMI. D) Representation of 4 of the 120 brain maps derived from automated meta-analyses of cognitive concepts included in the cognitive atlas, with color scale corresponding to z-statistic (see methods). The top four cognitive terms are shown, ranked via F- statistic of 120 post hoc ANOVA tests of the relationship between these maps and the Yeo Atlas. As the maps are largely symmetric, for illustrative purposes, the left hemisphere is shown for movement and working memory, while the right hemisphere is shown for autobiographic memory and pain. E) The Chi-square transformation of the MANOVA test statistic where the networks of the Yeo atlas were the dependent variable and the 120 cognitive maps were the independent variables, for the original data as well as the probability density distribution of 1000 rotational permutations. The P-value was calculated as the frequency with which the permuted Chi-square statistic equaled or exceeded the actual test statistic. F) The Chi-square transformation of the MANOVA test statistic where the networks of the Desikan atlas were the dependent variable and the 120 cognitive maps were the independent variables, for the original data as well as the probability density distribution of 1000 rotational permutations. The P-value was calculated as the frequency with which the permuted Chi-square statistic equaled or exceeded the actual test statistic.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Achard S, et al. Hubs of brain functional networks are radically reorganized in comatose patients. Proceedings of the National Academy of Sciences of the United States of America. 2012;109(50):20608–20613. - PMC - PubMed
    1. Alexander-Bloch A, et al. Subtle in-scanner motion biases automated measurement of brain anatomy from in vivo MRI. Human Brain Mapping. 2016;37(7):2385–2397. - PMC - PubMed
    1. Alexander-Bloch A, et al. The convergence of maturational change and structural covariance in human cortical networks. The Journal of Neuroscience. 2013;33(7):2889–2899. - PMC - PubMed
    1. Alexander-Bloch A, et al. The discovery of population differences in network community structure: new methods and applications to brain functional networks in schizophrenia. NeuroImage. 2012;59(4):3889–3900. - PMC - PubMed
    1. Alexander-Bloch AF, et al. Disrupted modularity and local connectivity of brain functional networks in childhood-onset schizophrenia. Frontiers in Systems Neuroscience. 2010;4:147. - PMC - PubMed

Publication types

MeSH terms

LinkOut - more resources