Elsevier

NeuroImage

Volume 46, Issue 1, 15 May 2009, Pages 39-46
NeuroImage

Technical Note
A probabilistic MR atlas of the human cerebellum

https://doi.org/10.1016/j.neuroimage.2009.01.045Get rights and content

Abstract

The functional organization of the cerebellum is reflected in large part by the unique afferent and efferent connectivity of the individual cerebellar lobules. This functional diversity on a relatively small spatial scale makes accurate localization methods for human functional imaging and anatomical patient-based research indispensable. Here we present a probabilistic atlas of the cerebellar lobules in the anatomical space defined by the MNI152 template. We separately masked the lobules on T1-weighted MRI scans (1 mm isotropic resolution) of 20 healthy young participants (10 male, 10 female, average age 23.7 yrs). These cerebella were then aligned to the standard or non-linear version of the whole-brain MNI152 template using a number of commonly used normalization algorithms, or to a previously published cerebellum-only template (Diedrichsen, J., 2006. A spatially unbiased atlas template of the human cerebellum. NeuroImage 33, 127–138.). The resulting average overlap was higher for the cerebellum-only template than for any of the whole-brain normalization methods. The probabilistic maps allow for the valid assignment of functional activations to specific cerebellar lobules, while providing a quantitative measure of the uncertainty of such assignments. Furthermore, maximum probability maps derived from these atlases can be used to define regions of interest (ROIs) in functional neuroimaging and neuroanatomical research. The atlas, made freely available online, is compatible with a number of widely used analysis packages.

Introduction

The nature of cerebellar information processing is one of the most intriguing problems in systems neuroscience. While the local circuitry is relatively homogenous across the cerebellar cortex, the input–output relationships of different parts of the cerebellum are diverse, and are key to understand its functional organization. The existence of several independent cortico-cerebellar loops has been posited (Desmond et al., 1997, Middleton and Strick, 1997, Ramnani, 2006). In particular, lobules V, VI, VIIb and VIIIa have reciprocal connections with primary motor cortex, whereas Crus II has connections with area 46 in prefrontal cortex (Kelly and Strick, 2003). Thus, a number of functionally diverse lobules collectively occupy a spatially small volume. This makes the accurate localization of functional and anatomical data to specific cerebellar lobules very important.

A detailed understanding of structure–function relationships in the cerebellum requires therefore a reliable standard anatomical reference for the human cerebellum. One important step was the landmark publication of a cerebellar MR atlas by Schmahmann et al. (2000). In this atlas, a T1-weighted image of a single human cerebellum was coregistered to the MNI (Montreal Neurological Institute) template (Evans et al., 1993) and carefully annotated, providing researchers with a clear consensus about terminology. The atlas is now widely used, for example to annotate results from fMRI group analysis with lobular labels.

This latter use of the atlas, however, carries the risk of systematic mis-assignments of anatomical labels to locations. The problem arises because of the relatively high spatial variability of individual cerebellar structures after alignment to the MNI template space. For example, after affine whole-brain alignment the primary fissures and the intrabiventer fissures of different individuals spread over an area of more than 1.5 cm in the common atlas space (Diedrichsen, 2006), an area as wide as the neighboring lobules themselves. Therefore, we cannot expect that the spatial arrangement of structures in an individual cerebellum is representative of the spatial arrangement in a population of participants.

For neocortical cytoarchitecture, the problem has been addressed through the development of a set of probabilistic atlases (Amunts and Zilles, 2001, Toga et al., 2006). Using post-mortem histological and chemical measures, different cytoarchitectonic areas were identified in individual brains and aligned into MNI reference space. Data from these individuals was combined, and the resulting maps indicate the proportion of subjects in whom a specific cytoarchitectonic field occupies a location in MNI reference space. This provides a measure of the likelihood that a particular coordinate in MNI space is occupied by cortex with certain cytoarchitectonic properties, and can be used to assign anatomical labels to functional activation in an informed, systematic and unbiased fashion (Eickhoff et al., 2007). The approach can also be used to define regions of interest to test a-priori hypotheses (Eickhoff et al., 2006).

Given the usefulness of probabilistic approaches to the anatomy of the neocortex, we thought to develop a similar tool for the human cerebellum. However, in the cerebellum a cyto-architectonic approach is not possible, as the local circuitry is homogenous, and we currently do not have reliable techniques to map the connectivity of the human cerebellum. Therefore, macro-anatomy currently provides the best source of information for classifying functionally distinct regions of the cerebellar cortex. In favor of a macro-anatomical approach is also that the subdivision of the cerebellar cortex into 10 lobules is relative invariant across normal healthy individuals and even across a range of mammalian species. Despite this invariance, there is inter-subject spatial variability in the ways that individual lobules occupy MNI reference space after registration. Here, we create a probabilistic representation of each of the cerebellar lobules and quantify this spatial variability.

In the development of probabilistic atlases, the choice of method used for registration or normalization of individual anatomies to the reference template is important. The spatial resolution of the normalization algorithm, the choice of reference template, and the involved preprocessing of the anatomical data will all influence the spatial variances of structures in reference space. Different normalization methods can potentially also lead to differences in the average location of structures in reference space, even if the same reference template is used. It is therefore important that the normalization method used to analyze the data matches as closely as possible the normalization method that is used to generate the atlas. To improve cerebellar alignment, we have recently developed a high-resolution template of the cerebellum and brainstem (SUIT, Diedrichsen, 2006) that considerably improves overlap compared to current whole-brain normalization methods. While the probabilistic atlas allows for the highest confidence of lobular assignments when using SUIT normalization, we also generated the atlas using a range of commonly used whole-brain normalization methods.

Section snippets

Participants

The structural images of twenty neurologically normal, healthy right-handed participants were used. There were ten male and ten female subjects, with ages ranging from 19 to 27, average 23.7 yrs. The racial composition of the sample reflects the diversity typical for fMRI studies with volunteers drawn from a University population: The sample included participants with Caucasian (14), East Asian (3), Indian (2), and African (1) racial origin. Structural images were acquired as part of two

Individual anatomy

Although individual lobules were clearly identifiable in each of the cerebella, the overall shape of the cerebellum in each subject varied markedly. This is illustrated in three individual cases in Fig. 1 (comparable coronal planes of section). For example, in some participants (e.g. Fig. 1A) lobule IX is pushed downward into the opening of the foramen magnum, whereas for the other participants (Figs. 1B, C) lobule IX is rolled inward and the neighboring lobule VIIIb forms the inferior surface

Discussion

The cerebellar atlas developed here relies on macro-anatomical landmarks to distinguish different lobules in the human cerebellum, following the current consensus in nomenclature (Schmahmann et al., 2000). This approach contrasts with recent probabilistic atlases of the human neocortex that are based on cytoarchitectonic information (Amunts and Zilles, 2001, Toga et al., 2006). While there are some chemical markers that indicate a compartmentalization of the cerebellum (Herrup and Kuemerle, 1997

Acknowledgments

The work was supported by Grants from the National Science Foundation (BSC 0726685, to JD), and the Biological Sciences Research Council (to NR). JHB was supported by RHUL-SGUL joint scholarship awarded to NR; EC supported by Welcome Trust Vacation Scholarship to NR and EC. We thank John Schlerf for helpful comments.

References (31)

  • MakrisN. et al.

    MRI-based surface-assisted parcellation of human cerebellar cortex: an anatomically specified method with estimate of reliability

    NeuroImage

    (2005)
  • SmithS.M. et al.

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

    NeuroImage

    (2004)
  • SnyderP.J. et al.

    Cerebellar volume asymmetries are related to handedness: a quantitative MRI study

    Neuropsychologia

    (1995)
  • AmuntsK. et al.

    Advances in cytoarchitectonic mapping of the human cerebral cortex

    Neuroimaging Clin. N. Am.

    (2001)
  • AnderssonJ.L. et al.

    FNIRT— FMRIB's non-linear image registration tool

    (2008)
  • Cited by (992)

    • Methods for cerebellar imaging analysis

      2023, Current Opinion in Behavioral Sciences
    View all citing articles on Scopus
    View full text