Elsevier

NeuroImage

Volume 33, Issue 1, 15 October 2006, Pages 127-138
NeuroImage

A spatially unbiased atlas template of the human cerebellum

https://doi.org/10.1016/j.neuroimage.2006.05.056Get rights and content

Abstract

This article presents a new high-resolution atlas template of the human, cerebellum and brainstem, based on the anatomy of 20 young healthy individuals. The atlas is spatially unbiased, i.e., the location of each structure is equal to the expected location of that structure across individuals in MNI space, a result that is cross-validated with an independent sample of 16 individuals. At the same time, the new template preserves the anatomical detail of cerebellar structures through a nonlinear atlas generation algorithm. In comparison to current whole-brain templates, it allows for an improved voxel-by-voxel normalization for functional MRI and lesion analysis. Alignment to the template requires that the cerebellum and brainstem are isolated from the surrounding tissue, a process for which an automated algorithm has been developed. Compared to normalization to the MNI whole-brain template, the new method strongly improves the alignment of individual fissures, reducing their spatial spread by 60%, and improves the overlap of the deep cerebellar nuclei. Applied to functional MRI data, the new normalization technique leads to a 5–15% increase in peak t values and in the activated volume in the cerebellar cortex for movement vs. rest contrasts. This indicates that the new template significantly improves the overlap of functionally equivalent cerebellar regions across individuals. The template and software are freely available as an SPM-toolbox, which also allows users to relate the new template to the annotated volumetric (Schmahmann, J.D., Doyon, J., Toga, A., Petrides, M., Evans, A. (2000). MRI atlas of the human cerebellum. San Diego: Academic Press) and surface-based (Van Essen, D.C. (2002a) Surface-based atlases of cerebellar cortex in the human, macaque, and mouse. Ann. N. Y. Acad. Sci. 978:468–479.) atlas of one individual, the “colin27”-brain.

Introduction

While the simple and homogenous cerebellar micro-circuitry suggests a uniform computational function of this structure, a unified theory of cerebellar function remains elusive. A number of hypotheses have been proposed, including the coordination of movement across different joints (Thach et al., 1992), timing (Ivry et al., 2002), internal models (Wolpert et al., 1998), or the cerebellum as a fast learning machine (Albus, 1971). With the advent of neuroimaging, it has also become apparent that parts of the cerebellum are involved in sensory (Gao et al., 1996), and cognitive processes (Courchesne and Allen, 1997).

Underlying this apparent functional heterogeneity is the fact that, while the cytoarchitecture of the cerebellum is homogenous, inputs and outputs are not: the cerebellum receives afferent fibers through the pons from nearly every cortical area (Schmahmann, 1996). These fibers appear to terminate in specialized regions of the cerebellar cortex and form closed loops with their respective cortical targets (Middleton and Strick, 1997, Kelly and Strick, 2003). While distinct functional areas in the cerebral cortex are of the size of multiple cm2, distinct subregions of the cerebellum may occur on a much smaller scale.

In applying functional magnetic resonance imaging (fMRI) to the cerebellum, the small scale of functionally distinct regions constitutes a major challenge. How does one combine anatomical and functional data across participants, given the considerable anatomical variability between individuals, and the rather small size of functional subunits? One possibility is to parcellate each cerebellum based on the individual anatomy into a set specific regions, e.g., individual lobules (Pierson et al., 2002, Makris et al., 2003, Makris et al., 2005). For functional analysis one would then average the BOLD signal within each region and subsequently average each region across individuals (Desmond et al., 1997, Desmond et al., 1998). Although such region-of-interest based approaches have been quite successful in single cases, they are not widely used because they are very labor intensive. More importantly, they limit analysis to a set of distinct set of subregions and do not allow for a more fine-grained voxel-based analysis.

Voxel-based approaches, in contrast, attempt a continuous mapping between the individual anatomy and a specific template (Woods et al., 1998b, Ashburner and Friston, 1999). In such approaches, the template image, g(x), is matched to an individual image, f(y), using the deformation map, yi = xi + v(xi), where x and y constitute locations in template and individual image, respectively. The deformation can be found by minimizing the cost function J, the squared voxel-by-voxel difference between the template g and thedeformed source image f:J=i(f(xi+v(xi))g(xi))2

In the simplest case, the deformation map v can be conceptualized as a 12-parameter affine transform (Woods et al., 1998b), yi = Axi + c, allowing for translation, rotation, scaling and shearing of the template space to map onto the individual's brain. For the cerebellum an affine alignment has been used by Grodd et al. (2001), only allowing for translation and scaling. This approach relied on a set of manually defined landmarks. Another popular approach, also utilized in this paper, uses cosine basis functions (Ashburner and Friston, 1999) to allow for nonlinear deformations (see Methods).

Currently, the most widely used template for voxel-based analysis is a template from the Montreal Neurological Institute (MNI), accepted as a standard by the International Consortium for Brain Mapping (ICBM). This template, the ICBM152, was generated by averaging 152 anatomical scans after correcting for overall brain size and orientation (Evans et al., 1993). As a result, the template provides very little anatomical detail. This, as we show below, leads to poor alignment of cerebellar structures, limiting the usefulness of a voxel-based approach for infra-tentorial structures.

To overcome these limitations, we aimed at developing an atlas template of the cerebellum and brainstem that represents the average geometry of a sample of individuals, while still providing enough anatomical detail to ensure that individual lobules within the cerebellum can be aligned. Such a template should therefore improve the overlap of cerebellar structures, while retaining the advantages of a voxel-based approach (Woods et al., 1998b, Ashburner and Friston, 1999).

One possible solution for a template would have been to use the cerebellum of a particular individual, for example colin27, a young individual who was scanned 27 times at the Montreal Neurological Institute. The cerebellar anatomy of this individual has been carefully documented (Schmahmann et al., 2000), and a flattened representation of this cerebellum has been created (Van Essen, 2002b). However, using a single individual's anatomy as a template, as was done by Talairach and Tournoux (1988), has an important drawback: every individual shows some anatomical idiosyncrasies that are not representative of the population. When normalizing a sample of individual brains to this space, systematic deformations would arise, which could affect both functional and anatomical studies of the cerebellum.

Therefore, our goal was to make a spatially unbiased template (for a similar argument, see Woods et al., 1998a). With this we mean that the location of any particular structure i in the new atlas template should be equal to the average, or expected, location of that structure across all individuals n:E(yi(n))=zii

Spatial bias can only be defined in respect to a common reference frame in which locations across individuals can be compared. As a commonly accepted reference frame, we chose here the ICBM152 template. Therefore, the new cerebellar template was generated using a group of 20 participants that had undergone an affine alignment to the ICBM152 whole-brain template. To limit the template and the normalization to the cerebellum and brainstem, we developed an algorithm that isolates the cerebellum and brainstem from the surrounding tissue. This ensures that the boundaries of the cerebellum are properly aligned across individuals. After isolation, all cerebella underwent a nonlinear normalization to a single individual cerebellum, and were averaged. This average image was then deformed using the inverse average deformation, creating a new Spatially Unbiased Infra-tentorial (SUIT) template. As a result, the coordinates of a structure in the new template are equal to the average coordinates of that structures across individuals (Eq. (2)). At the same time, by using the isolation algorithm and a high-resolution nonlinear deformation, the new template preserves much more anatomical detail than the whole-brain template.

We show in an independent cross-validation sample of 16 participants that the new atlas template leads to a significantly improvement in overlap of individual cerebellar fissures and of the deep cerebellar nuclei. We also show that the template significantly improves the analysis of functional data, making the new template a useful tool for functional and lesion analyses of the human cerebellum.

Section snippets

Participants

The atlas is based on anatomical data from 20 neurologically healthy subjects (11 females and 9 males). Their ages ranged from 22 to 45 years, the mean age was 27.25 years. The atlas group was comprised of 13 participants from Exp 1 in Diedrichsen et al. (2005) and 7 new participants. For cross-validation we used the 16 participants (6 females and 10 males, ages 18–29 years, mean age 23. 8 years) that had participated in Exp 2 of the same study. The Johns Hopkins School of Medicine Internal

Isolation of cerebellar and brainstem

Isolation of the cerebellum and brainstem from surrounding tissue took approximately 4 min per individual on a standard PC-laptop machine. A successful example of resulting probability maps is shown in Fig. 3A. The outline of the cerebellum is clearly defined and the map could be used without manual correction for the subsequent normalization steps. In the majority of the individuals, however, it was necessary to manually correct the segmentation because of frequently occurring defects. In many

Discussion

We present here a cerebellar atlas template that is based on the average geometry of a group of individuals, and at the same time preserves the detail of cerebellar anatomy. The atlas template was created such that it is spatially unbiased, i.e., the location of structures in the template represents the expected location of the corresponding structures in the individual anatomies after affine normalization to the ICBM152 template. Thus, average coordinates within the new SUIT template can be

Acknowledgments

This work was supported through grants from the NIH (NS37422), the Human Frontiers Science Program, and the Johns Hopkins General Clinical Research Center (RPN 02-08-15-03). I am grateful to the staff of the F.M. Kirby Research Center for Functional Brain Imaging at Kennedy Krieger Institute, funded by an NIH/NCRR resource grant (RR15241), and to Scott Grafton, Reza Shadmehr, and Sarah Ying for the helpful comments and support.

References (35)

  • J. Ashburner et al.

    Nonlinear spatial normalization using basis functions

    Hum. Brain Mapp.

    (1999)
  • E. Courchesne et al.

    Prediction and preparation, fundamental functions of the cerebellum

    Learn. Mem.

    (1997)
  • J.E. Desmond et al.

    Lobular patterns of cerebellar activation in verbal working-memory and finger-tapping tasks as revealed by functional MRI

    J. Neurosci.

    (1997)
  • J. Diedrichsen et al.

    Neural correlates of reach errors

    J. Neurosci.

    (2005)
  • A. Dimitrova et al.

    Probabilistic 3D MRI atlas of the human cerebellar dentate/interposed nuclei

    NeuroImage

    (2005)
  • A.C. Evans et al.

    3D statistical neuroanatomical models from 305 MRI volumes

    Proc. IEEE-Nucl. Sci. Symp. Med. Imaging Conf.

    (1993)
  • B. Fischl et al.

    High-resolution intersubject averaging and a coordinate system for the cortical surface

    Hum. Brain Mapp.

    (1999)
  • Cited by (704)

    • Methods for cerebellar imaging: cerebellar subdivision

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