Elsevier

NeuroImage

Volume 80, 15 October 2013, Pages 397-404
NeuroImage

The parcellation-based connectome: Limitations and extensions

https://doi.org/10.1016/j.neuroimage.2013.03.053Get rights and content

Highlights

  • Parcellation schemes may be used to examine the macroscopic connectome.

  • The choice of parcellation scheme has an effect on topological characteristics.

  • Qualitative descriptions are robust, but precise numerical values fluctuate.

  • Results from different studies are therefore not directly comparable.

  • Advances in parcellation techniques and network tools may improve comparability.

Abstract

The human connectome is an intricate system of interconnected elements, providing the basis for integrative brain function. An essential step in the macroscopic mapping and examination of this network of structural and functional interactions is the subdivision of the brain into large-scale regions. Parcellation approaches used for the formation of macroscopic brain networks include application of predefined anatomical templates, randomly generated templates and voxel-based divisions. In this review, we discuss the use of such parcellation approaches for the examination of connectome characteristics. We specifically address the impact of the choice of parcellation scheme and resolution on the estimation of the brain's topological and spatial network features. Although organizational principles of functional and structural brain networks appear to be largely independent of the adopted parcellation approach, quantitative measures of these principles may be significantly modulated. Future parcellation-based connectome studies might benefit from the adoption of novel network tools and promising advances in connectivity-based parcellation approaches.

Introduction

The brain is a complex network of structurally and functionally interlinked regions. Integrative brain function does not solely depend on the activation or influence of a single element of this network, but rather emerges from the system of interactions as a whole, the human connectome (Sporns et al., 2005). Inspired by pioneering work on the structure of animal connectomes, such as those of the nematode Caenorhabditis elegans (White et al., 1986), cat (Scannell et al., 1995, Scannell et al., 1999) and primate (Felleman and van Essen, 1991, Young, 1993), recent advances in neuroimaging techniques (Basser et al., 2000, Lowe et al., 2000, Tuch, 2004, Wedeen et al., 2005) are utilized to map the brain's functional and structural connections in ever increasing detail (Cammoun et al., 2012, Gong et al., 2009, Hagmann et al., 2007, Toga et al., 2012, Tuch, 2004, Wedeen et al., 2005). Studies examining the spatial and topological organization of these connectome maps have revealed several characteristics of an efficient network architecture, including aspects of economic wiring (Bassett et al., 2010, Bullmore and Sporns, 2012), the presence of clustered functional communities, indicative of information segregation (Bullmore and Sporns, 2009, Damoiseaux et al., 2006, Hilgetag and Kaiser, 2004, Hilgetag et al., 2000, Salvador et al., 2005, van den Heuvel et al., 2009a), short communication relays (Hagmann et al., 2007, Kaiser and Hilgetag, 2006, Sporns and Zwi, 2004, van den Heuvel et al., 2009b, van den Heuvel et al., 2012, Varshney et al., 2011) and the formation of a small set of densely connected and centrally placed hub nodes, playing a crucial role in global brain communication (Achard et al., 2006, Collin et al., 2013, Hagmann et al., 2008, Sporns et al., 2007, van den Heuvel and Sporns, 2011, van den Heuvel et al., 2012).

Due to the vast amounts of neurons and the small measurement scale, reconstructing an entire connectome on the neuronal level, be it of a simple life form as the Drosophila fly (Chiang et al., 2011, Chklovskii et al., 2010) or zebrafish (Stobb et al., 2012), let alone that of mammalian species, is an ambitious and challenging task. According to estimates, the human brain consists of circa 100 billion neurons, their soma varying between 4 and 100 μm in diameter, and a thousandfold connections (Azevedo et al., 2009, Williams and Herrup, 1988). For the mapping and analysis of this microscopic connectome, further advances in both imaging techniques (e.g., higher spatial resolutions and shorter acquisition times) and computer systems (increasing data storage and computational power) are required. Today, a complete connectome map on the microscopic scale is only available for the simple nematode C. elegans, with a nervous system consisting of just 302 neurons (Varshney et al., 2011, White et al., 1986). Connectome maps of the human brain are instead formed on a macroscopic scale, mapping structural and functional connections between large-scale brain regions rather than interactions between single neurons.

To allow a rescaling of the connectome from a microscopic to a macroscopic scale, a definition of the brain's large-scale regions and connections is required. The process of dividing the brain into such macroscopic regions is usually referred to as “parcellation”. In particular when it comes to the delineation of structurally or functionally distinct regions in the cerebral cortex, there is a long research tradition (see Zilles and Amunts (2010) and Geyer et al. (2011) for an overview), going back more than a century to the seminal work of Penfield (Penfield, 1950, Penfield and Boldrey, 1937) and Brodmann (Brodmann, 1909, Brodmann, 1914). Through the ages, cortical areas have been distinguished on basis of, e.g., cytoarchitecture (Brodmann, 1909, Schleicher et al., 1999, von Economo and Koskinas, 1925), myeloarchitecture (Vogt and Vogt, 1919), electrophysiological observations (Lim et al., 1994, Penfield and Boldrey, 1937), immunohistochemistry (Baleydier et al., 1997) and receptor autoradiography (Zilles et al., 2002). In recent years, novel techniques based on magnetic resonance imaging (MRI) have extended this list, exploiting cortical curvature patterns (Desikan et al., 2006, Van Essen, 2005), in vivo mappings of myelin content (Glasser and Van Essen, 2011) and similarities in structural and functional connectivity profiles (Craddock et al., 2012, Johansen-Berg et al., 2004). Results from these different parcellation criteria have been brought together in multiple three-dimensional brain atlases, also referred to as anatomical templates, which can be used to parcellate a subject's brain on basis of an MRI scan (Collins et al., 1995, Fischl et al., 2004, Tzourio-Mazoyer et al., 2002). Together with high-resolution parcellation approaches that subdivide the brain into voxels (Eguíluz et al., 2005, van den Heuvel et al., 2008a) or randomly generated parcels (Hagmann et al., 2007, van den Heuvel and Sporns, 2011), these brain atlases are at the heart of parcellation-based connectome analysis, providing the large-scale brain regions whose structural and functional connections are the topic of investigation.

For the examination of topological and spatial features of the brain's wiring pattern, a formal description of the system as a whole is needed. To this end, a connectome (map) is often represented as a graph; a mathematical notion of a network, describing elements (nodes) and their interactions (edges) in an abstract manner (Bullmore and Sporns, 2009), allowing a systematic study of the network's topology through computation of so-called graph metrics (Fig. 1) (Rubinov and Sporns, 2010, van den Heuvel et al., 2010). In the graph representation of a parcellation-based connectome, each node of the graph corresponds to a brain region from the template that was used to subdivide the brain. In turn, the edges of the graph correspond to structural or functional connections between these brain regions, depending on the utilized imaging modality. Since the boundaries and amount of distinguished regions, ranging in size from single voxels to large-scale parcels of hundreds of voxels, vary considerably between templates, the adopted parcellation approach directly implicates the (number of) nodes and therefore also the number and placement of edges. That is, the formal representation of a parcellation-based connectome depends on the choice of parcellation scheme.

Here, we review the application of parcellation templates for brain networks established by diffusion tractography (Basser et al., 2000, Hagmann et al., 2006, Mori et al., 1999, Wedeen et al., 2005) and resting-state functional magnetic resonance imaging (fMRI) (Biswal et al., 1995, Fox et al., 2005, Lowe et al., 2000) and discuss their impact on the estimation and examination of connectome characteristics. Special attention will be paid to the effect of parcellation resolution (i.e., the number of regions in a template), which has recently been suggested to play an important role in the assessment of network features. Following the main discussion on the use and impact of parcellation templates, we briefly touch upon the exciting field of connectivity-based parcellation, which aims to discern the fundamental functional and structural elements of the human brain on basis of the connectivity profile of brain regions (Anwander et al., 2007, Johansen-Berg et al., 2004).

Section snippets

Predefined anatomical templates

A common approach to select the nodes of a macroscopic brain network is to parcellate the brain of a given subject using a predefined anatomical template or brain atlas. Typically, the subject's anatomical images are registered to a template brain whose regions have been labeled using one or more manually parcellated subjects from a fixed training set. The labels of this template brain are then propagated onto the registered anatomical images, thus subdividing the subject's gray matter into

Influence on network metrics

The influence of spatial scale on macroscopic connectome characteristics has been addressed in several studies. Most importantly, all show a strong impact of network resolution on graph metrics. Bassett et al. (2011) repeatedly bisected the regions of the AAL, HO and LPBA40 atlases, obtaining templates of two, four and eight times the original resolutions. Similar to the comparison between anatomical templates, architectural principles were found to be robust across spatial scales, but network

Connectivity-based parcellation

Connectivity-based parcellation approaches are an exciting and upcoming class of parcellation techniques that aim to delineate large-scale brain regions by exploiting similarities in structural or functional connectivity patterns (Cloutman and Lambon Ralph, 2012). Similar to connectome research, connectivity-based parcellation is motivated by the idea that brain function may be understood from the underlying wiring pattern (Johansen-Berg et al., 2004, Sporns et al., 2005). Rather than

Conclusions

An important aspect of the mapping and analysis of parcellation-based connectomes is the choice of parcellation scheme, which directly implicates the nodes and edges of a macroscopic brain network. In this review, we have reflected upon the use and impact of several parcellation approaches, most prominently the application of predefined anatomical templates and random high-resolution templates. Studies examining the influence of parcellation schemes all support the same conclusion:

Conflict of interest

The authors have no conflict of interest to declare.

References (111)

  • J.H. Kim

    Defining functional SMA and pre-SMA subregions in human MFC using resting state fMRI: functional connectivity-based parcellation method

    NeuroImage

    (2010)
  • J.C. Klein et al.

    Connectivity-based parcellation of human cortex using diffusion MRI: establishing reproducibility, validity and observer independence in BA 44/45 and SMA/pre-SMA

    NeuroImage

    (2007)
  • S. Lim et al.

    Functional anatomy of the human supplementary sensorimotor area: results of extraoperative electrical stimulation

    Electroencephalogr. Clin. Neurophysiol.

    (1994)
  • M.J. Lowe et al.

    Correlations in low-frequency BOLD fluctuations reflect cortico-cortical connections

    NeuroImage

    (2000)
  • G. Robins et al.

    An introduction to exponential random graph (p*) models for social networks

    Soc. Netw.

    (2007)
  • M. Rubinov et al.

    Complex network measures of brain connectivity: uses and interpretations

    NeuroImage

    (2010)
  • A. Schleicher et al.

    Observer-independent method for microstructural parcellation of cerebral cortex: a quantitative approach to cytoarchitectonics

    NeuroImage

    (1999)
  • D.W. Shattuck et al.

    Construction of a 3D probabilistic atlas of human cortical structures

    NeuroImage

    (2008)
  • S.L. Simpson et al.

    An exponential random graph modeling approach to creating group-based representative whole-brain connectivity networks

    Neuroimage

    (2012)
  • S.M. Smith

    The future of FMRI connectivity

    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)
  • M.P. van den Heuvel et al.

    Exploring the brain network: a review on resting-state fMRI functional connectivity

    Eur. Neuropsychopharmacol.

    (2010)
  • M.P. van den Heuvel et al.

    Small-world and scale-free organization of voxel-based resting-state functional connectivity in the human brain

    Neuroimage

    (2008)
  • D.C. Van Essen

    A population-average, landmark- and surface-based (PALS) atlas of human cerebral cortex

    Neuroimage

    (2005)
  • S. Achard et al.

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

    J. Neurosci.

    (2006)
  • R. Albert et al.

    Statistical mechanics of complex networks

    Rev. Mod. Phys.

    (2002)
  • L.A.N. Amaral et al.

    Classes of small-world networks

    Proc. Natl. Acad. Sci.

    (2000)
  • A. Anwander et al.

    Connectivity-based parcellation of Broca's area

    Cereb. Cortex

    (2007)
  • F.A. Azevedo et al.

    Equal numbers of neuronal and nonneuronal cells make the human brain an isometrically scaled-up primate brain

    J. Comp. Neurol.

    (2009)
  • C. Baleydier et al.

    Neurofilament architecture of superior and mesial premotor cortex in the human brain

    NeuroReport

    (1997)
  • P.J. Basser et al.

    In vivo fiber tractography using DT-MRI data

    Magn. Reson. Med.

    (2000)
  • D.S. Bassett et al.

    Efficient physical embedding of topologically complex information processing networks in brains and computer circuits

    PLoS Comput. Biol.

    (2010)
  • 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)
  • J.W. Bohland et al.

    The brain atlas concordance problem: quantitative comparison of anatomical parcellations

    PLoS One

    (2009)
  • K. Brodmann

    Vergleichende Lokalisationslehre der Grosshirnrinde

    (1909)
  • K. Brodmann

    Physiologie des Gehirns

    (1914)
  • R.L. Buckner et al.

    Cortical hubs revealed by intrinsic functional connectivity: mapping, assessment of stability, and relation to Alzheimer's disease

    J. Neurosci.

    (2009)
  • E. Bullmore et al.

    Complex brain networks: graph theoretical analysis of structural and functional systems

    Nat. Rev. Neurosci.

    (2009)
  • E. Bullmore et al.

    The economy of brain network organization

    Nat. Rev. Neurosci.

    (2012)
  • G. Cecchi et al.

    Identifying directed links in large scale functional networks: application to brain fMRI

    BMC Cell Biol.

    (2007)
  • P. Christie et al.

    The interpretation and application of Rent's rule

    IEEE Trans. Very Large Scale Integr. (VLSI) Syst.

    (2000)
  • L.L. Cloutman et al.

    Connectivity-based structural and functional parcellation of the human cortex using diffusion imaging and tractography

    Front. Neuroanat.

    (2012)
  • G. Collin et al.

    Structural and Functional Aspects Relating to Cost and Benefit of Rich-Club Organization in the Human Cerebral Cortex

    (2013)
  • D.L. Collins et al.

    Automatic 3-D model-based neuroanatomical segmentation

    Hum. Brain Mapp.

    (1995)
  • L.D.F. Costa et al.

    Beyond the average: detecting global singular nodes from local features in complex networks

    Europhys. Lett.

    (2009)
  • R.C. Craddock et al.

    A whole brain fMRI atlas generated via spatially constrained spectral clustering

    Hum. Brain Mapp.

    (2012)
  • J.S. Damoiseaux et al.

    Consistent resting-state networks across healthy subjects

    Proc. Natl. Acad. Sci.

    (2006)
  • C. Echtermeyer et al.

    Integrating temporal and spatial scales: human structural network motifs across age and region-of-interest size

    Front. Neuroinform.

    (2011)
  • C. Echtermeyer et al.

    Automatic network fingerprinting through single-node motifs

    PLoS One

    (2011)
  • Cited by (235)

    • Towards a biologically annotated brain connectome

      2023, Nature Reviews Neuroscience
    View all citing articles on Scopus
    View full text