Introduction
The cerebellum is a functionally highly diverse structure: Each lobules has its unique pattern of connectivity with the neocortex, and therefore has likely its own functional role. To promote accurate anatomical inferences for human functional and anatomical imaging studies, we present here a probabilistic atlas of the cerebellar lobules in the space defined by the MNI152 template. The anatomical definitions are based on the fMRI atlas of an individual cerebellum by Schmahmann et al. (2000). To obtain a representative anatomical atlas, we separately masked the lobules on T1-weighted MRI scans (1mm isotropic resolution) of 20 healthy young participants (10 male, 10 female, average age 23.7 yrs). Using a different set of 23 particpants, we also masked the deep cerebellar nucelei. These cerebella were then aligned using different commonly used normalization algorithms.
The resultant probabilistic maps allow for the valid assignment of functional activations to specific cerebellar lobules and the nuclei, while providing a quantitative measure of the certainty 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 (more on this here). The atlas is included in the newer releases of FSL and the Anatomy toolbox. More version of the atlases for use with MRICroN are also available here.
Please cite the atlas as:
- Diedrichsen J., Balster J.H., Flavell J., Cussans E., Ramnani N. (2009). A probabilistic MR atlas of the human cerebellum. Neuroimage.
- Diedrichsen J., Maderwald S., Küper M., Thürling M., Rabe K., Gizewski ER, Ladd M, Timmann D (under review). Imaging the deep cerebellar nuclei: A probabilistic atlas and normalization procedure. Neuroimage.
The generation of the atlas was supported by a Grant by the National Science foundation (NSF) - BSC 0726685.
Why a probabilistic atlas?
 |
It is common practice to align individual brains using the non-linear normalization in SPM, and then to assign foci in the group analysis to lobules using the atlas by Schmahmann et al. (2000). Our analysis shows that these assignments are only correct for 38%(!) of the voxels: In about 2/3 of the voxels the atlas is a least one lobule off. There are two reasons for this:
These problems stress the importance of a representative atlas and of good correspondence between the normalization method used for the atlas and for the group analysis. For this reason we provide the atlas generated with a number of different normalization methods. |
The probabilistic atlas also provides a measure of the certainty with with anatomical assignments can be made. The image shows the proportion of the 20 individuals that overlapped with the same lobulus in atlas space. We generated the atlas for:
The overlap in SUIT (A) often reaches 100% whereas the overlap using affine methods (FLIRT, B) is somewhat poorer. |  |
 |
The left figure shows the percent overlap between different lobules after normalization. Newer non-linear methods (Segmentation in SPM, FNIRT) lead to a good correspondence and to relatively high accuracies. Cerebellum-only normalization with SUIT leads to the best overlap. Thus, for new imaging studies of the cerebellum we would strong recommend one of the newer non-linear methods, or ideally the use of SUIT. For the interpretation of older results (for example for meta-analyses of cerebellar imaging results), we make the probabilistic atlas for the SPM nonlinear normalization also available. |
In collaboration with Prof. Timmann from the University of Duisburg-Essen we also imaged the deep cerebellar nuclei in a set of 23 separate subjects at 7T, using susceptibility-weighted imaging. The dentate, emoliform, globose and fastigial nuclei relay all output from the cerebellum and are therefore a extremely important part of the cerebellar circuit. In the most recent version of the probabilistic atlas we provide probabilistic ROIs for the dentate, the interposed (emboliform and globose) and fastigial nuclei. |
 |
The Atlas for FSL View
 |
The atlas can be used with the Atlas Widget in FSLView, part of the FSL package. The atlas for FSLView is available for
The FLIRT and FNIRT versions are already included in newer distributions of FSL. The lobular assignment is approximately spatially unbiased in these atlases, however the maximal probabilities after SUIT alignment is in general higher.
|
The Atlas for MRICroN
|
The atlas can be used MRICroN, which is an image viewer written by Chris Rorden and which is freely available for Linux, Windows, and Mac OS X.The atlas for MRICroN is available for:
The necessary files (Maximum-probabiliy-map, size of maximum probability, full probability maps, color map (lut) and text files are all included. To use MRICroN to look up probabilities
|
The Atlas for Anatomy Toolbox in SPM
|
The atlas is now included in the Anatomy toolbox, which was developed S. Eickhoff and colleagues. For the generation of the maps, we used the Segementation and Normalisation algorithm in SPM5/8. To make the maps fit with the rest of the toolbox, they were then warped into the space defined by the Colin-brain. |
Registration and Download
Note that the SUIT-MRICro version of the probabilistc atlas comes with the new release (v. 2.4) of the SUIT toolbox.
