Surface-Based Display of Volume-Averaged Cerebellar Imaging Data.
Diedrichsen J, Zotow E. · 2015
Read Surface-Based Display of Volume-Averaged Cerebellar Imaging Data. by Diedrichsen J, Zotow E. (2015) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4521932.txt.
Introduction
Jörn Diedrichsen, Ewa Zotow, and Izumi Sugihara present a flat map of the cerebellum in PLoS ONE in 2015. The cerebellum, at the back of the brain, is folded so tightly that standard volume displays hide much of its surface. Scans averaged across people have lacked a surface view like the ones long used for the cortex.
Their method builds white- and grey-matter surfaces from the averaged anatomy itself, rather than from each person's brain. Scan data are projected onto those surfaces, then unfolded into a flatmap sized to match the true grey-matter area. The whole surface shows at once, with front-to-back and side-to-side order readable. Example maps show rest networks and task activity, and the tools are freely shared.
Catalog record
| CREATOR | Diedrichsen J, Zotow E. |
|---|---|
| YEAR | 2015 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Surface-Based Display of Volume-Averaged Cerebellar Imaging Data By Jörn Diedrichsen, Ewa Zotow, Izumi Sugihara Published in: PLoS ONE DOI: 10.1371/journal.pone.0133402 ABSTRACT Abstract The paper presents a flat representation of the human cerebellum, useful for visualizing functional imaging data after volume-based normalization and averaging across subjects. Instead of reconstructing individual cerebellar surfaces, the method uses a white- and grey-matter surface defined on volume-averaged anatomical data. Functional data can be projected along the lines of corresponding vertices on the two surfaces. The flat representation is optimized to yield a roughly proportional relationship between the surface area of the 2D-representation and the volume of the underlying cerebellar grey matter. The map allows users to visualize the activation state of the complete cerebellar grey matter in one concise view, equally revealing both the anterior-posterior (lobular) and medial-lateral organization. As examples, published data on resting-state networks and task-related activity are presented on the flatmap. The software and maps are freely available and compatible with most major neuroimaging packages. INTRODUCTION Surface-based analysis and visualization methods have greatly contributed to our understanding of the functional organization of the neocortex [1–3]. Traditional volume-based displays provide only a particular view of the data, and even adjacent slices through the volume can often tell a very different visual story. Inflated representations of the cortical sheet allow for the visualization of functional activity patterns across the entire neocortex. Surface-based methods also allow for improved accuracy of inter-subject alignment, accounting for individual variability in the structure of cortical folding [4–8]. While surface-based analysis and visualization have become a gold standard in many laboratories and software packages, they are not yet used for data from the cerebellum. There are two reasons for this. First, the cerebellum is much more tightly folded than the human neocortex, with individual folia being 1-2mm wide [9] and the complete cerebellar cortical sheet approximately 1.5-2m long [10]. Complete unfolding of the cerebellar cortex based on MRI scans requires a resolution higher than 200μm [11], which is typically not available in human imaging studies. Secondly, for the currently typical resolution of functional imaging data (2-4mm), the activation from adjacent folia is averaged. Therefore, even small and focal activations cannot unambiguously be assigned to a specific folium, reducing the precision that could be gained by reconstructing individual surfaces. Thus, instead of developing a surface-based normalization for individually reconstructed cerebellar surfaces, the aim of this paper is to provide a flat representation of the cerebellum as a visualization tool for volume-averaged cerebellar data. Volume-based alignment of cerebellar data has improved greatly through the development of non-linear morphing algorithms and high-resolution non-linear templates [12–15]. These methods superimpose individual cerebellar lobules with good precision [16], even if they cannot bring individual folia into alignment. A reconstruction of an individual cerebellar surface has been published previously [2]. This map reflects the folding of the cerebellum to the level of groups of folia of a single individual, as accurately as is possible for an anatomical scan with a 1mm resolution. For display of functional group data, however, this surface is unsatisfactory. For example, in the volume-based analysis, there are two sites of individual finger representations in the human cerebellum [17]. However, when projecting this data onto the aforementioned flatmap, both loci are fractionated and distributed over a number of folds of the surface (Fig 1a). This problem arises from the fact that the level of detail of the anatomical unfolding is too high for the resolution of volume-averaged group data. [Figure: Representation of individual fingers in the human cerebellum.Shown is the classification accuracy with which the moved finger can be determined from the local pattern of activity, with a threshold of z>1 [17]. (A) Data projected onto a surface based on a single anatomy [2] displays single closed activation clusters as a fractured series of blobs. (B) Projection to the new flatmap ensures that single clusters in the volume are also presented as such on the surface.] Therefore, we aimed here to develop a surface representation of the body of cerebellar grey matter that meets the following criteria: It should show data on the level of lobules, but not attempt to unfold the cerebellar surface down to the level of individual folia. The projection should ensure that closed clusters in the volume are also displayed as closed clusters on the surface (Fig 1b). The projection method should ensure that the entire data from cerebellar grey matter is represented on the surface. Lastly, the 2D-map should provide a truthful representation of the size of different cerebellar structures with the surface area corresponding approximately to the displayed volume. After presenting and evaluating the surface representation, we demonstrate the utility of the flatmap as a tool for visualizing various kinds of cerebellar data. We first project the probabilistic atlas of the lobular structure [12] to the flatmap, which provides a first basic anatomical orientation. We then present maps of functional resting-state connectivity obtained from a large data set (N = 1000) [18], which reveal the intricate pattern of connectivity with the neocortex. Finally, we present task-based activation maps using a subsample (N = 100) of participants from the Human Connectome Project (HCP) [19], who were scanned on a variety of functional tasks. While the results from these data sets have been analyzed and published for the neocortex [20], the authors did not present the cerebellar findings in full detail. Together, the presented data sets should provide a useful reference relative to which new results can be interpreted. METHOD Data and ethics statement Anatomical and functional data from five published studies was utilized [12, 16–18, 20]. All experimental procedures were approved by the respective local ethics committees, as documented in the original papers. The re-analysis of these non-sensitive and fully anonymized data did not require separate approval under the rules of the University College London research ethics committee. Volume preparation The flatmap was based on the anatomical data from 20 participants originally used for the generation of the spatially unbiased infratentorial template (SUIT) [12]. The scans were segmented into grey and white matter using unified segmentation [13] and the cerebellum was isolated from the neocortex using an automatic algorithm [12]. After the first affine alignment to the SUIT template, we utilized the fast diffeomorphic anatomical registration algorithm (Dartel) [14] to generate a new, slightly sharper average grey-matter and white-matter template. The extent of the cerebellar white-matter body was then estimated applying a threshold of p>0.4 to the average map of white-matter probabilities. The remaining cerebellar voxels were labeled as grey matter if the grey matter probability exceeded 0.5. The maps were then hand-edited to ensure that the grey matter of the posterior vermis was separated from the abutting hemispheres of lobule HIX. Surface reconstruction We reconstructed the outer grey-matter surface (Fig 2a) and the inner white-matter surface (Fig 2b) using the Freesurfer packages [1]. After smoothing and topology correction, both surfaces were inflated. Twenty-two pairs of reference points were then selected, which defined an approximate mapping between the vertices of the two surfaces. These points were placed on clearly defined landmarks, such as the base and superficial aspect of major fissures. Each vertex from the inflated outer surface was then projected to the inflated inner surface, using the connection