Publications
Text queries can be conducted by Author, Title, or Keyword.
High resolution cerebellar anatomy
Source: 2000;.
Author: Toga AW, Holmes CJ.
Abstract:
Jansen and Brodal began their treatise on the cerebellum by pointing out its great morphological diversity across species, which appeared striking even within mammals (Jansen and Brodal, 1954). While intriguing the early anatomists, this fascinating anatomy presents considerable challenges to current methods for imaging, mapping and measuring morphology. These tasks are further complicated by today's focus on functional imaging, which requires that the brain be mapped in vivo. The cerebellum's gross features and major landmarks are easily distinguished with non-invasive techniques such as conventional magnetic resonance imaging (MRI), but novel techniques are required to discern its individual folia and deep nuclei. While the creation of stereotaxic atlas systems for the cerebral hemispheres has greatly facilitated the exhange and comparison of structural and functional data, the most ubiquitous sterotaxic systems and atlas spaces fail to sufficiently define the cerebellum in terms of its placement or delineation. This chapter describes progress in each of these problem areas. Specifically, we describe the use of a high resolution cryosectioning approach that produces full color 3D image volumes of in situ anatomy and a multi-scan MRI approach to achieve superior in vivo image volumes of cerebellar anatomy. We also describe efforts to rectify the standard cerebral atlases with multisubject mappings of this structure by use of informatics techniques and a deformable brain atlas. CRYOSECTIONING: Several years ago, we sought to create a description of brain anatomy that surpassed the resolution of in vivo techniques but retained the tomographic and 3D attributes of in situ sampling. We also wanted to collect data that had sufficiently high resolution and retained the texture and color available to histologically treated tissue. The product of our efforts was a method of digitally acquiring images from the block face of cryosectioned whole brain. This enabled reconstructions of unparalleled resolution in full color of the entire 3D volume. The cerebellum was sampled along with the cerebrum and as a benefit of a high sampling frequency of isotropic pixels, could be sliced in any orientation favorable to the question at hand. These methods allow the collection of serially sectioned full-color images of brain within and without the cranium using a high-resolution camera. Digitization of the block face itself during the ongoing sectioning process preserves the spatial integrity of the data volume and reduces the time to produce comprehensive reconstructions. The results yield a 3D volume. The spatial resolution of images derived from surface photography of sectioned anatomic specimens is higher than the limits of conventional MR imaging and can be increased either by the use of higher pixel count instruments, or the capture of smaller fields of view. The combination of cryosectioning and specimen surface photography provides the means for acquiring anatomic image data and the potential for the simultaneous collection of specimen tissue for histological analysis. Such methods have been used to yield high-resolution, morphologically detailed imagery for correlation with other imaging modalities such as MR and CT or histology. Cryoplaned specimens also have been used in the production of digital atlases along with conventional film photography taken from the surface of the preparations. However, such atlases based on secondary reconstruction of non-registered serial images suffer from alignment and sampling problems. HIGH-RESOLUTION DIGITAL ANATOMY: We were able to appreciate the natural color of the non-perfused brain. The color quality demonstrated in these full color images provided subtle texture and contrast characteristics that were not as readily apparent in comparison to similar monochrome images. Since these images of the block face are not histologically processed, the blood that remained in the non-perfused, fresh frozen tissue facilitated gray/white matter discrimination and made apparent many of the observed subnuclei and other smaller structures. The embedding medium used to fix the brain en bloc to the cryotome stage was colored to facilitate distinction between brain and background. This was especially beneficial when examining the cerebellum because of the relatively smaller folia and interstitial spaces. This difference was especially appreciated when edited data were compared with original digital images that contained the supporting ice block. Collection of the cerebellum in this way allowed for the observation of both lobular and deep nuclear anatomy. From the peduncles to the tips of the folia, the cerebellar white matter was clearly visible. Frequently, penetrating vessels appeared within the fibers as puncta or as dark blue striations. The larger fissures were clearly visible as the lobules generally separated slightly along these structures. The remaining finer fissures were also clearly determined, whereas in some cases intrafoliar divisions were more difficult to visualize, especially in the depths where the tissue was tightly pressed together and partial volume effects became problematic. The deep nuclei of the cerebellum were evident, with the dentate being most easily observed in any plane. The fastigial and emboliform were less well distinguished, however, and the cell clusters forming the globus nucleus were quite difficult to make out. Associated structures in the brainstem, such as the olivary nucleus of the pons, were also finely discriminated in the cryosection volumes. The results of these post mortem experiments produced unparalleled anatomic detail of the cerebellum while retaining tomographic 3D spatial relationships. Further, the cerebellar anatomy was studied relative to the rest of the brain, and hence was amenable to modern brain mapping and atlasing strategies. These atlas strategies require a complete brain volume for correct placement within a coordinate system, within which comparisons between different modalities of the same subject can be made. IN VIVO HUMAN CERBELLAR ANATOMY: There are a variety of robust computational methods for the registration of 3D volumes between and within subjects and scanning modalities. In our approach, these tools are applied to intramodality scans from the same subject, with the result being post hoc enhancement of the signal-to-noise ratio through averaging. As each scan can be kept short, this gain comes without the cost of long individual scan times and therefore allows the production of higher quality images than those available from single MR scanning protocols. To evaluate this process, we scanned a single subject multiple times, compared the result of averaging these scans to single high resolution scans, and examined the usefulness of the resulting volume in other brain-mapping applications. The resulting collection of T1-weighted scans were registered and averaged from 27 different scanning sessions of the same individual. The product of this was vastly superior imagery describing the anatomy of the whole brain including the cerebellum. The averaging reduced the noise inherent in MRI and resulted in better edge detection as well as discrimination of subnuclei. This intramodality, intrasubject averaging also helped the quality of T2 and proton dense images from the same sessions. Coupled with the cryosectioned data, these approaches provide a unique advantage for describing cerebellar anatomy and when placed within a common reference system can be used to interpret traditional MRI scans from other subjects. In contrast to standard clinical MR imaging, the averaging technique revealed not only the larger fissures, but the fine divisions between and within the folia. Additionally, enhancing the contrast of the image allowed the dentate nucleus to be visualized in vivo. DEFORMABLE BRAIN ATLASES: In view of the complex structural variability between individuals, a fixed digital atlas, representing the anatomy of a single human brain, will fail to serve as a faithful representation of the brains of new subjects. It would, however, be ideal if an atlas could be elastically deformed to fit a new image set from an incoming subject. Transforming individual datasets into the shape of a single reference anatomy, or onto a 3D digital brain atlas, removes subject-specific shape variations, and allows subsequent comparison of brain function between individuals. Conversely, high-dimensional warping algorithms can also be used to transfer all the information in a 3D digital brain atlas onto the scan of any given subject, while respecting the intricate patterns of structural variation in their anatomy. Such deformable atlases can be used to carry 3D maps of functional and vascular territories into the coordinate system of different subjects, as well as information on different tissue types and the boundaries of cytoarchitectonic fields and their neurochemical composition. Deformable atlases rely on high-dimensional warping algorithms to drive them into precise structural correspondence with target brain images. The fine external features of the cerebellum make it ideally suited to a surface-based warping procedure, recently devised and implemented in our laboratory. This algorithm was designed to calculate the high-dimensional deformation field relating the brain anatomies of an arbitrary pair of subjects or within the same subject over time, and to transfer functional information between subjects or integrate that information on a single anatomic template. High spatial accuracy can be guaranteed by using a large set of corresponding anatomic surfaces to constrain the complex transformation of one of one subject’s anatomy into the shape of another. These surfaces include critical functional interfaces such as the vermis and hemispheres, as well as numerous cytoarchitectonic and lobar boundaries in 3 dimensions. Connected systems of parametric meshes model deep internal fissures, or sulci, lobules or other readily observable boundaries whose trajectories represent critical functional divisions. These sulci are sufficiently extended inside the brain to reflect subtle and distributed variations in neuroanatomy between subjects. The parametric form of the system of connected surface elements allows us to represent the relation between any pair of anatomies as a family of high-resolution displacement maps carrying the surface systems of one individual onto another in stereotaxic space. The algorithm then calculates the high-dimensional volumetric warp deforming one 3D scan into structural correspondence with the other. Integral distortion functions are used to extend the deformation field required to elastically transform these surface systems into structural correspondence with their counterparts in the target scan. 3D warping algorithms provide a method for calculating local and global shape changes and give valuable information about normal and abnormal growth and development. Deformable atlases not only account for the anatomic variations and idiosyncrasies of each individual subject, but they offer a powerful strategy for exploring and classifying age-related, developmental or pathologic variations in anatomy. More fundamentally, they also provide a method for spatially normalizing the anatomies of different brains. While the more esoteric of these algorithms may lie outside the reach of smaller laboratories, the development of distributed computing resources and the automation of warping algorithms may ultimately allow remote users to submit their own images for normalization to regional or national computational centers. Such a facility for consistent cross-center normalization and registration could supply a much more robust basis for comparing experimental or clinical data obtained from different subject or different research centers than exists at the present time. CONCLUSION: The use of novel methods to describe the anatomy of the cerebellum coupled with sophisticated atlas and warping approaches makes the study of this structure considerably more tractable. Increasing the spatial resolution of the source data provides obvious advantages in distinguishing details of the anatomy. However, most techniques that provide cyto- or chemoarchitectural information do so destructively and without benefit of 3D whole brain reference. The cryosectioning and MRI averaging methods described here offer unique bridging modalities that help extend detailed neuroanatomic information into a multisubject deformable atlas framework.
Source: 2000;.
Author: Toga AW, Holmes CJ.
Abstract:
Jansen and Brodal began their treatise on the cerebellum by pointing out its great morphological diversity across species, which appeared striking even within mammals (Jansen and Brodal, 1954). While intriguing the early anatomists, this fascinating anatomy presents considerable challenges to current methods for imaging, mapping and measuring morphology. These tasks are further complicated by today's focus on functional imaging, which requires that the brain be mapped in vivo. The cerebellum's gross features and major landmarks are easily distinguished with non-invasive techniques such as conventional magnetic resonance imaging (MRI), but novel techniques are required to discern its individual folia and deep nuclei. While the creation of stereotaxic atlas systems for the cerebral hemispheres has greatly facilitated the exhange and comparison of structural and functional data, the most ubiquitous sterotaxic systems and atlas spaces fail to sufficiently define the cerebellum in terms of its placement or delineation. This chapter describes progress in each of these problem areas. Specifically, we describe the use of a high resolution cryosectioning approach that produces full color 3D image volumes of in situ anatomy and a multi-scan MRI approach to achieve superior in vivo image volumes of cerebellar anatomy. We also describe efforts to rectify the standard cerebral atlases with multisubject mappings of this structure by use of informatics techniques and a deformable brain atlas. CRYOSECTIONING: Several years ago, we sought to create a description of brain anatomy that surpassed the resolution of in vivo techniques but retained the tomographic and 3D attributes of in situ sampling. We also wanted to collect data that had sufficiently high resolution and retained the texture and color available to histologically treated tissue. The product of our efforts was a method of digitally acquiring images from the block face of cryosectioned whole brain. This enabled reconstructions of unparalleled resolution in full color of the entire 3D volume. The cerebellum was sampled along with the cerebrum and as a benefit of a high sampling frequency of isotropic pixels, could be sliced in any orientation favorable to the question at hand. These methods allow the collection of serially sectioned full-color images of brain within and without the cranium using a high-resolution camera. Digitization of the block face itself during the ongoing sectioning process preserves the spatial integrity of the data volume and reduces the time to produce comprehensive reconstructions. The results yield a 3D volume. The spatial resolution of images derived from surface photography of sectioned anatomic specimens is higher than the limits of conventional MR imaging and can be increased either by the use of higher pixel count instruments, or the capture of smaller fields of view. The combination of cryosectioning and specimen surface photography provides the means for acquiring anatomic image data and the potential for the simultaneous collection of specimen tissue for histological analysis. Such methods have been used to yield high-resolution, morphologically detailed imagery for correlation with other imaging modalities such as MR and CT or histology. Cryoplaned specimens also have been used in the production of digital atlases along with conventional film photography taken from the surface of the preparations. However, such atlases based on secondary reconstruction of non-registered serial images suffer from alignment and sampling problems. HIGH-RESOLUTION DIGITAL ANATOMY: We were able to appreciate the natural color of the non-perfused brain. The color quality demonstrated in these full color images provided subtle texture and contrast characteristics that were not as readily apparent in comparison to similar monochrome images. Since these images of the block face are not histologically processed, the blood that remained in the non-perfused, fresh frozen tissue facilitated gray/white matter discrimination and made apparent many of the observed subnuclei and other smaller structures. The embedding medium used to fix the brain en bloc to the cryotome stage was colored to facilitate distinction between brain and background. This was especially beneficial when examining the cerebellum because of the relatively smaller folia and interstitial spaces. This difference was especially appreciated when edited data were compared with original digital images that contained the supporting ice block. Collection of the cerebellum in this way allowed for the observation of both lobular and deep nuclear anatomy. From the peduncles to the tips of the folia, the cerebellar white matter was clearly visible. Frequently, penetrating vessels appeared within the fibers as puncta or as dark blue striations. The larger fissures were clearly visible as the lobules generally separated slightly along these structures. The remaining finer fissures were also clearly determined, whereas in some cases intrafoliar divisions were more difficult to visualize, especially in the depths where the tissue was tightly pressed together and partial volume effects became problematic. The deep nuclei of the cerebellum were evident, with the dentate being most easily observed in any plane. The fastigial and emboliform were less well distinguished, however, and the cell clusters forming the globus nucleus were quite difficult to make out. Associated structures in the brainstem, such as the olivary nucleus of the pons, were also finely discriminated in the cryosection volumes. The results of these post mortem experiments produced unparalleled anatomic detail of the cerebellum while retaining tomographic 3D spatial relationships. Further, the cerebellar anatomy was studied relative to the rest of the brain, and hence was amenable to modern brain mapping and atlasing strategies. These atlas strategies require a complete brain volume for correct placement within a coordinate system, within which comparisons between different modalities of the same subject can be made. IN VIVO HUMAN CERBELLAR ANATOMY: There are a variety of robust computational methods for the registration of 3D volumes between and within subjects and scanning modalities. In our approach, these tools are applied to intramodality scans from the same subject, with the result being post hoc enhancement of the signal-to-noise ratio through averaging. As each scan can be kept short, this gain comes without the cost of long individual scan times and therefore allows the production of higher quality images than those available from single MR scanning protocols. To evaluate this process, we scanned a single subject multiple times, compared the result of averaging these scans to single high resolution scans, and examined the usefulness of the resulting volume in other brain-mapping applications. The resulting collection of T1-weighted scans were registered and averaged from 27 different scanning sessions of the same individual. The product of this was vastly superior imagery describing the anatomy of the whole brain including the cerebellum. The averaging reduced the noise inherent in MRI and resulted in better edge detection as well as discrimination of subnuclei. This intramodality, intrasubject averaging also helped the quality of T2 and proton dense images from the same sessions. Coupled with the cryosectioned data, these approaches provide a unique advantage for describing cerebellar anatomy and when placed within a common reference system can be used to interpret traditional MRI scans from other subjects. In contrast to standard clinical MR imaging, the averaging technique revealed not only the larger fissures, but the fine divisions between and within the folia. Additionally, enhancing the contrast of the image allowed the dentate nucleus to be visualized in vivo. DEFORMABLE BRAIN ATLASES: In view of the complex structural variability between individuals, a fixed digital atlas, representing the anatomy of a single human brain, will fail to serve as a faithful representation of the brains of new subjects. It would, however, be ideal if an atlas could be elastically deformed to fit a new image set from an incoming subject. Transforming individual datasets into the shape of a single reference anatomy, or onto a 3D digital brain atlas, removes subject-specific shape variations, and allows subsequent comparison of brain function between individuals. Conversely, high-dimensional warping algorithms can also be used to transfer all the information in a 3D digital brain atlas onto the scan of any given subject, while respecting the intricate patterns of structural variation in their anatomy. Such deformable atlases can be used to carry 3D maps of functional and vascular territories into the coordinate system of different subjects, as well as information on different tissue types and the boundaries of cytoarchitectonic fields and their neurochemical composition. Deformable atlases rely on high-dimensional warping algorithms to drive them into precise structural correspondence with target brain images. The fine external features of the cerebellum make it ideally suited to a surface-based warping procedure, recently devised and implemented in our laboratory. This algorithm was designed to calculate the high-dimensional deformation field relating the brain anatomies of an arbitrary pair of subjects or within the same subject over time, and to transfer functional information between subjects or integrate that information on a single anatomic template. High spatial accuracy can be guaranteed by using a large set of corresponding anatomic surfaces to constrain the complex transformation of one of one subject’s anatomy into the shape of another. These surfaces include critical functional interfaces such as the vermis and hemispheres, as well as numerous cytoarchitectonic and lobar boundaries in 3 dimensions. Connected systems of parametric meshes model deep internal fissures, or sulci, lobules or other readily observable boundaries whose trajectories represent critical functional divisions. These sulci are sufficiently extended inside the brain to reflect subtle and distributed variations in neuroanatomy between subjects. The parametric form of the system of connected surface elements allows us to represent the relation between any pair of anatomies as a family of high-resolution displacement maps carrying the surface systems of one individual onto another in stereotaxic space. The algorithm then calculates the high-dimensional volumetric warp deforming one 3D scan into structural correspondence with the other. Integral distortion functions are used to extend the deformation field required to elastically transform these surface systems into structural correspondence with their counterparts in the target scan. 3D warping algorithms provide a method for calculating local and global shape changes and give valuable information about normal and abnormal growth and development. Deformable atlases not only account for the anatomic variations and idiosyncrasies of each individual subject, but they offer a powerful strategy for exploring and classifying age-related, developmental or pathologic variations in anatomy. More fundamentally, they also provide a method for spatially normalizing the anatomies of different brains. While the more esoteric of these algorithms may lie outside the reach of smaller laboratories, the development of distributed computing resources and the automation of warping algorithms may ultimately allow remote users to submit their own images for normalization to regional or national computational centers. Such a facility for consistent cross-center normalization and registration could supply a much more robust basis for comparing experimental or clinical data obtained from different subject or different research centers than exists at the present time. CONCLUSION: The use of novel methods to describe the anatomy of the cerebellum coupled with sophisticated atlas and warping approaches makes the study of this structure considerably more tractable. Increasing the spatial resolution of the source data provides obvious advantages in distinguishing details of the anatomy. However, most techniques that provide cyto- or chemoarchitectural information do so destructively and without benefit of 3D whole brain reference. The cryosectioning and MRI averaging methods described here offer unique bridging modalities that help extend detailed neuroanatomic information into a multisubject deformable atlas framework.