Elsevier

Journal of Neuroscience Methods

Volume 254, 30 October 2015, Pages 46-59
Journal of Neuroscience Methods

Basic neuroscience
AICHA: An atlas of intrinsic connectivity of homotopic areas

https://doi.org/10.1016/j.jneumeth.2015.07.013Get rights and content

Highlights

  • AICHA a functional brain ROIs atlas based on resting-state fMRI data acquired in 281 individuals.

  • AICHA include a crucial aspect of cerebral organization, namely homotopy.

  • Each region having a unique homotopic contralateral counterpart with which it has maximal intrinsic connectivity.

  • AICHA is ideally suited for investigating brain hemispheric specialization.

Abstract

Background

Atlases of brain anatomical ROIs are widely used for functional MRI data analysis. Recently, it was proposed that an atlas of ROIs derived from a functional brain parcellation could be advantageous, in particular for understanding how different regions share information. However, functional atlases so far proposed do not account for a crucial aspect of cerebral organization, namely homotopy, i.e. that each region in one hemisphere has a homologue in the other hemisphere.

New method

We present AICHA (for Atlas of Intrinsic Connectivity of Homotopic Areas), a functional brain ROIs atlas based on resting-state fMRI data acquired in 281 individuals. AICHA ROIs cover the whole cerebrum, each having 1—homogeneity of its constituting voxels intrinsic activity, and 2—a unique homotopic contralateral counterpart with which it has maximal intrinsic connectivity. AICHA was built in 4 steps: (1) estimation of resting-state networks (RSNs) using individual resting-state fMRI independent components, (2) k-means clustering of voxel-wise group level profiles of connectivity, (3) homotopic regional grouping based on maximal inter-hemispheric functional correlation, and (4) ROI labeling.

Results

AICHA includes 192 homotopic region pairs (122 gyral, 50 sulcal, and 20 gray nuclei). As an application, we report inter-hemispheric (homotopic and heterotopic) and intra-hemispheric connectivity patterns at different sparsities.

Comparison with existing method

ROI functional homogeneity was higher for AICHA than for anatomical ROI atlases, but slightly lower than for another functional ROI atlas not accounting for homotopy.

Conclusion

AICHA is ideally suited for intrinsic/effective connectivity analyses, as well as for investigating brain hemispheric specialization.

Introduction

Over the previous ten years, the use of brain parcellation atlases (Evans et al., 2012) has become more frequent with the introduction of resting state functional magnetic resonance imaging (fMRI) connectivity analysis (Biswal et al., 1995). Within this framework, atlas based sets of regions of interest (ROI) are used to compute a brain graph, i.e., a model of the human brain functional connectome (Bullmore and Bassett, 2011). As shown by Craddock et al. (2012), region-based analysis has advantages compared with voxel-based analysis, including better sensitivity, interpretability and computational time. Furthermore, the reduction of the dimensionality of the fMRI data makes the problem statistically manageable by lowering the required number of statistical tests (Zalesky et al., 2012).

Today, several atlases are available, and each atlas has specific properties that should be considered with respect to the user's needs. The available atlases for functional/intrinsic connectivity and graph analyses can be characterized by the dataset of images that served in their elaboration, the target space that had been chosen for normalization, the neuro-anatomo/functional parameters that drive the parcellation, and the parcellation algorithm.

The datasets involved in atlas creation define their “specific versus generic” nature. At one end are the most “specific” atlases, that are built from a single-subject dataset (such as in the AAL case (Tzourio-Mazoyer et al., 2002); at the other end are the most “generic” atlases, that are built from datasets representative of the general population. Note that elaboration of a truly “generic” atlas is out of reach, since it would request datasets for all ages, ethnicities, etc. Rather, available generic atlases are based on samples representative of some specific population, balancing for some given phenotypes, such as sex (Craddock et al., 2012), or conversely, selecting a specific phenotype, such right-handedness (Shen et al., 2013).

Regarding the target space, two main choices are available, namely volume (Ashburner and Friston, 2005) or surface (Dale et al., 1999). While the surface space has been demonstrated to be more accurate (Jo et al., 2007), the volume space is important when one is interested in tissue other than the cortical gray matter mantle, such as white matter or nuclei. The partial or complete parcellation proposed for a given brain tissue class should also be considered in atlas selection. For example, network centrality graph analysis (Bullmore and Bassett, 2011) can be critically affected in the case of partial coverage atlases because such analyses consider the strength of the connection between all graph regions. For example, centrality analysis based on an incomplete parcellation atlas or a complete cortex coverage atlas can lead to discrepant sets of “hub” region identification.

For each atlas, there are some neuro-anatomo-functional parameters that guide the parcellation scheme and constrain the definition of landmarks that constitute the boundaries of the atlas regions. Three types of landmarks are used for gray matter parcellation, including sulci (anatomical landmarks, Tzourio-Mazoyer et al., 2002, Kennedy et al., 1998), borders of cytoarchitectonic areas (Caspers et al., 2006) or limits defined from functional properties, such as the borders of spatially coherent regions of homogeneous functional connectivity (Craddock et al., 2012). Note that the homogeneity in the regions of interest of a given atlas will depend on the parameters that guide its parcellation scheme.

The atlas building constraints that are explicitly or implicitly associated with the parcellation scheme, such as the choice of the landmarks, has a strong impact on the topology of the parcellation in terms of the shape and number of parcels and thus on the atlas’ final resolution.

The aim of the present work was to design an atlas suitable for functional/intrinsic connectivity and graph analysis that would consider and benefit from a major characteristic of brain organization, namely homotopy. Homotopy corresponds to the fact that the two hemispheres have comparable organization regarding macroscopic anatomy, cytoarchitecture and large-scale functional organization (Fuster, 1998, Mesulam, 1990) and are for the most part connected to each other by the white matter tracts of the corpus callosum, which originate from and terminate in the 4th cortical layer. As a consequence, almost all areas from identical cortical structures and with identical hierarchical levels are connected through the corpus callosum in a mirroring way. However, this cortical symmetry is not perfect because of the global torsion of the brain, the Yaklovian torque, which makes difficult a point-to-point correspondence between cortical areas that are functionally homotopic (Toga and Thompson, 2003). In addition, the torque goes along with asymmetries in sulcus depth and position in relation to differences in asymmetries in neighboring cortices (Lyttelton et al., 2009); the largest sulcal and cortical asymmetries in newborns and adults are located at the termination of the Sylvian fissure and at the superior temporal sulcus (Hill et al., 2010). These gross morphological differences across hemispheres increase the difficulty in defining homotopic regions in the temporal and inferior parietal areas. For example, in order to calculate regional asymmetries, one must define what is meant by homotopic regions. This definition can be based either on anatomical atlases, such as AAL (Tzourio-Mazoyer et al., 2002), which use a rough spatial metric based on the position of the regions relative to the sulci, or on a cytoarchitectonic atlas based on both anatomical symmetry in position and differences in the cortical lamination patterns (Caspers et al., 2006). In functional atlases, criteria based on the functional characteristics of the regions of the atlas areas have not yet been proposed to define homotopic areas. However, it is common to observe homotopic-like patterns in resting data analysis using seed based analysis (van den Heuvel and Hulshoff Pol, 2010, Jo et al., 2012) or network decomposition methods (Beckmann et al., 2005, Naveau et al., 2012, Yeo et al., 2011). Furthermore, Stark et al. (2008) have reported that the highest mean functional connectivity is observed between inter-hemispheric symmetrical areas compared with other inter-hemispheric or intra-hemispheric connections. These observations support that homotopic organization is a fundamental feature of the functional organization of the cortex and plead for the definition of homotopic areas based on functional, rather than anatomical, criteria.

The Atlas of Intrinsic Connectivity of Homotopic Areas (AICHA) we propose here is thus a population-level, cerebrum gray matter brain atlas of homotopic regions based on a time correlation structure of the resting brain. We designed AICHA based on the following properties: 1—a large dataset of 281 healthy participants balanced for handedness and gender; 2—volumetric MNI space of normalization to cover the gray matter of the whole cerebrum, including the sub-cortical gray nuclei; 3—time correlation structure of the functional resting state signal (intrinsic connectivity) as the basis for landmark definition; and 4—choice of an algorithm that permits a homotopic parcellation to maximize the weight of the functional signal at the final step. We describe the AICHA atlas building, its dataset regional functional homogeneity (i.e., the intra-regional similarity level of the voxel blood oxygen level dependent time-series), compare it to four existing atlases, and provide the intra- and inter- hemispheric intrinsic connectivity patterns at different thresholds of regional connection strength.

Section snippets

Participants

Two hundred eighty-one healthy adults (137 women, 144 men) aged 18–57 years (25 ± 6 years, mean ± SD) were included in this study. Subject handedness was self-reported, and their manual preference strength was measured using the Edinburgh score (ES, mean ± σ, (Oldfield, 1971). 133 subjects were right-handed (66 women, ES = 94 ± 10; 67 men, ES = 91 ± 13) and 148 subjects were left-handed (71 women, ES = −63 ± 39; 77 men, ES = −63 ± 41). All subjects provided informed written consent, and the local ethics committee

General description of AICHA ROIs

AICHA includes 192 pairs of homotopic ROIs (Fig. 3). Of these ROIs, 122 ROIs anatomically belong to 37 gyri (Table 1), 50 ROIs overlap 14 sulci (Table 2) and 20 ROIs are localized within the 4 subcortical gray matter nuclei (Table 2). The complete labels are provided as Supplementary material (see Supplementary Fig. S5), together with the ROI volumes and mass center coordinates. The average volume of the AICHA ROIs is 2.1 ± 2 cm3 (mean ± S.D.) with a range from 0.16 to 10.2 cm3. A three-dimensional

Discussion

We will discuss the choices we made in the construction of AICHA, including the advantages and potential limitations of these choices.

Conclusion

We have developed an atlas of ROIs that covers the entire cerebral gray matter using a parcellation scheme based on the intrinsic connectivity of homotopic areas. Considering that AICHA was developed from resting state low frequency BOLD pseudo-oscillatory activity, which is known to subtend intrinsic brain function and likely task performance, we believe that this atlas will be useful for connectivity and graph analysis of both resting-state and task-related functional imaging data. In

Acknowledgments

We thank Guy Perchey for his essential help with data acquisition.

References (75)

  • D.G. Gee et al.

    Low frequency fluctuations reveal integrated and segregated processing among the cerebral hemispheres

    NeuroImage

    (2011)
  • M.F. Glasser et al.

    The minimal preprocessing pipelines for the Human Connectome Project

    NeuroImage

    (2013)
  • Y. Golland et al.

    Data-driven clustering reveals a fundamental subdivision of the human cortex into two global systems

    Neuropsychologia

    (2008)
  • P.Y. Herve et al.

    Handedness and cerebral anatomical asymmetries in young adult males

    NeuroImage

    (2006)
  • P.Y. Herve et al.

    Revisiting human hemispheric specialization with neuroimaging

    Trends Cogn Sci

    (2013)
  • P.Y. Hervé et al.

    Revisiting human hemispheric specialization with neuroimaging

    Trends Cogn Sci

    (2013)
  • J. Himberg et al.

    Validating the independent components of neuroimaging time series via clustering and visualization

    NeuroImage

    (2004)
  • H.J. Jo et al.

    Spatial accuracy of fMRI activation influenced by volume- and surface-based spatial smoothing techniques

    NeuroImage

    (2007)
  • O.C. Lyttelton et al.

    Positional and surface area asymmetry of the human cerebral cortex

    NeuroImage

    (2009)
  • R.C. Oldfield

    The assessment and analysis of handedness: the Edinburgh inventory

    Neuropsychologia

    (1971)
  • J. Sacher et al.

    Sexual dimorphism in the human brain: evidence from neuroimaging

    Magn Reson Imaging

    (2013)
  • G. Salimi-Khorshidi et al.

    Automatic denoising of functional MRI data: combining independent component analysis and hierarchical fusion of classifiers

    NeuroImage

    (2014)
  • X. Shen et al.

    Groupwise whole-brain parcellation from resting-state fMRI data for network node identification

    NeuroImage

    (2013)
  • S.M. Smith et al.

    Advances in functional and structural MR image analysis and implementation as FSL

    NeuroImage

    (2004)
  • Z.Y. Sun et al.

    The effect of handedness on the shape of the central sulcus

    NeuroImage

    (2012)
  • N. Tzourio-Mazoyer et al.

    Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain

    NeuroImage

    (2002)
  • G. Varoquaux et al.

    A group model for stable multi-subject ICA on fMRI datasets

    NeuroImage

    (2010)
  • M. Vigneau et al.

    Meta-analyzing left hemisphere language areas: phonology, semantics, and sentence processing

    NeuroImage

    (2006)
  • A. Zalesky et al.

    Connectivity differences in brain networks

    NeuroImage

    (2012)
  • S. Achard et al.

    A resilient, low-frequency, small-world human brain functional network with highly connected association cortical hubs

    J Neurosci

    (2006)
  • J. Ashburner et al.

    Nonlinear spatial normalization using basis functions

    J Acoust Soc Am

    (1999)
  • C.F. Beckmann et al.

    Investigations into resting-state connectivity using independent component analysis

    Philos Trans R Soc Lond, B: Biol Sci

    (2005)
  • B. Biswal et al.

    Functional connectivity in the motor cortex of resting human brain using echo-planar MRI

    Magn Reson Med

    (1995)
  • K. Brodmann et al.

    Brodmann's ‘localisation in the cerebral cortex’

    (1994)
  • E.T. Bullmore et al.

    Brain graphs: graphical models of the human brain connectome

    Annu Rev Clin Psychol

    (2011)
  • V.D. Calhoun et al.

    A method for making group inferences from functional MRI data using independent component analysis

    Hum Brain Mapp

    (2001)
  • R.C. Craddock et al.

    A whole brain fMRI atlas generated via spatially constrained spectral clustering

    Hum Brain Mapp

    (2012)
  • Cited by (188)

    View all citing articles on Scopus
    View full text