SUIT is 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 close to the expected location of that structure across individuals in MNI space. At the same time, the template preserves the anatomical detail of cerebellar structures through a nonlinear atlas-generation algorithm. By using automated nonlinear normalization methods, a more accurate intersubject-alignment than current whole-brain methods can be achieved. Additionally, a probabilistic atlas of lobules and the deep cerebellar nuclei is available.
The template and software are freely available as an open source SPM-toolbox. The toolbox allows you to:
- Automatically isolate cerebellar structures from the cerebral cortex based on an anatomical image
- Achieve accurate anatomical normalisation of cerebellar structures
- Normalize functional imaging data for fMRI group analysis
- Display the functional data on a surface-based flatmap representation
- Normalize focal cerebellar lesions for lesion-symptom mapping
- Use Voxel-based morphometry (VBM) to determine patterns of cerebellar degeneration or growth
- Use a probabilistic atlas in SUIT space to assign locations to different cerebellar lobules in an unbiased and informed way
- Automatically define ROIs for specific cerebellar lobules and summarize function and anatomical data
- Improve normalization of the deep cerebellar nuclei using an ROI-driven normalization.
- The new version uses Dartel (developed by John Ashburner) as a normalization engine for more accurate results.
Please cite the use of this template as:
- Diedrichsen, J. (2006). A spatially unbiased atlas template of the human cerebellum. Neuroimage, 33, 1, p. 127-138.
- Diedrichsen, J., Balsters, J. H., Flavell, J., Cussans, E., & Ramnani, N. (2009). A probabilistic atlas of the human cerebellum. Neuroimage.
- Diedrichsen, J., Maderwald, S., Kuper, M., Thurling, M., Rabe, K., Gizewski, E. R., et al. (2011). Imaging the deep cerebellar nuclei: A probabilistic atlas and normalization procedure. Neuroimage.
- Diedrichsen, J. & Zotow, E. (2015). Surface-based display of volume-averaged cerebellar data. PLoS One, 7, e0133402.
The toolbox has been developed by:
J. Diedrichsen and improved with help of C. Hernandez-Castillo, G. Prichard, N. Lally, T.Wiestler, J. Schlerf, and many others.
Why a specialized template?
The most commonly used atlas template for brain imaging is the ICBM152 template, defining the so-called MNI space. The template was generated by averaging 152 individual brains after affine normalization (correcting for size, translation and rotation). As it can be seen on the left, the template provides very little contrast for cerebellar structures.
Normalization (affine or nonlinear) to this template leads to a large spatial spread of individual fissures. One the right is depicted the superposition of 16 individual primary fissures (A) and intra-biventer fissures (B). The distance between identical fissures in two different individuals is 4 mm on average. In the neo-cortex the large variability of the folding pattern prevents a more accurate alignment. For infra-tentorial structures (cerebellum and brainstem) the spatial structural variability is much less and would allow for a much more accurate alignment.
We therefore developed a new template of the cerebellum and brainstem, the spatially unbiased infra-tentorial template (SUIT). This template is:
Spatial unbiasedness, and the relationship to MNI space
The SUIT template was developed to be spatially unbiased in respect to the affine alignment to MNI space. This means that when normalizing the same brain to the MNI whole-brain template and to the SUIT template, the same structure should end up on average at the same coordinate in atlas space. For each individual there will be of course differences between the two normalizations (otherwise what would be the point?). These differences can be as big as 1 cm, and are on average across the image ~5mm. Furthermore, small systematic differences between different MNI normalisation methods (SPM Segmentation, SUIT, FSL's FNIRT) still remain. Therefore for the most accurate results, it is important that your data and your atlas template are normalised with the same algorithm. For the surface-based representation and the probabilistic atlas, we have therefore released separate versions for different atlas spaces.
The newest version of the toolbox also includes a surface-based representation of the SUIT-template. This allows you to display volume-averaged cerebellar data on a group flatmap - providing an intuitive and efficient way to communicate cerebellar activation results.
A few notes of warning:
- The new isolation only works using SPM12, the previous version of suit_isolate is still included for backward compatibility.
- The "Normalization" option in SPM2-SPM8 leads to substantial stretching of the cerebellum in the z-direction. Thus, coordinate in SUIT and the normalization methods are not comparable with "MNI" coordinates obtained using the non-linear normalization in SPM.
- Other normalization algorithms - normalization with concurrent segmentation (SPM5/8), FLIRT (FSL), and SUIT are reasonable unbiased in respect to each other. However, small differences exists, so we have released different versions for the probabilistic atlas and the cerebellar surface for different normalisation methods.
- The anatomy of the Colin cerebellum has been carefully documented in a MRI atlas of the human cerebellum (Schmahmann, 2002). Although this individual has been aligned to the MNI space, it is not unbiased, unlike our template. This means, it is not valid to take MNI-space coordinates and refer to the Colin brain to assign a lobule to this coordinate. As a solution, we have developed a probabilistic atlas of the human cerebellum.
- While the code is written to be compatible with SPM5-12 , all of our testing happens in SPM8/12 and we cannot guarantee that newer features will work under SPM5. SPM2 is no longer supported.