A deformable digital brain atlas system according to Talairach and Tournoux
Introduction
Neurosurgical interventions have to be carefully planned with high precision to allow maximally sparing trajectories to brain lesions (Kelly, 1996). All frequently used imaging modalities, especially magnetic resonance imaging (MRI), support the physician in identifying important structures in the brain and bearing them in mind for the surgical strategy. However, many of the brain structures are hardly or not at all visible in the acquired images. Therefore, neurosurgeons frequently consult brain atlases during the planning of interventions to improve their orientation. Brain atlases are usually printed books (the most well known are those of Schaltenbrand and Wahren (1977) and Talairach and Tournoux (1988)), which makes their application to patient’s images tedious and often erroneous. Because there are differences between an atlas book and a MR image concerning dimensionality (2D plates versus 3D volume dataset), medium of representation (printed on paper versus digital display), and anatomic shape (standard anatomy versus individual brain), the information transfer from the atlas to the MR image has to happen solely in the mind of the surgeon. It is evident that this procedure stresses the physician’s 3D imaginative capability very much, and it requires a long time experience to gain success. The usage of a book in the aseptic environment of the operating theatre is an additional problem which limits the application of an atlas to the preoperative planning stage. An other disadvantage of a printed atlas book is its finality, i.e. there is no way to correct errors and include new insights or further information.
To overcome these problems, as neurosurgeons demand, we decided to develop a computerized atlas system. As anatomic basis we chose the well-established stereotaxic brain atlas of Talairach and Tournoux (1988), although we know that the Talairach atlas itself has several disadvantages: it is two-dimensional (2D), is relatively sparse due to inter-slice distances of 2–5 mm, assumes left/right symmetry, and contains some inconsistencies between orthogonal plates. However, one can not neglect the fact that it is actually used by many neurosurgical groups for planning of interventions. Certainly this is the reason why in the past several working groups have built computerized atlas systems around the Talairach atlas. But, as we will discuss later, all systems proposed so far have some shortcomings.
We state a list of requirements which our computerized atlas system should fulfill in order to overcome the disadvantages of printed atlas books as well as those of the Talairach atlas and those of other existing systems:
- •
It shall include 3D reconstructed surface models of brain structures as well as the original atlas plates,
- •
it shall provide an easy-to-handle matching feature to adapt the atlas nonrigidly to individual brain images,
- •
it shall offer a powerful visualization to display atlas and MRI in joint views,
- •
it shall include additional information which extends the contents of the Talairach atlas; moreover it shall be open to further extensions, and
- •
it shall offer an interface to a navigation system in order to provide the atlas information intraoperatively.
Our paper is organized as follows. In Section 2 we give a short survey over related work. After this, we present the methods we applied for developing a computerized atlas system in Section 3. In particular, we discuss our 3D reconstruction algorithm (Section 3.1) as well as all the methods necessary for achieving nonrigid matching (Section 3.2). In Section 4, we describe the computerized Talairach atlas system which we developed in respect to accuracy considerations as well as to some application cases. Finally, we discuss the whole development in Section 5, where we also outline further extensions.
Section snippets
Related work
Many groups work on the development of computerized brain atlases. Several reports on atlases based on proprietary sources of anatomic knowledge have been published in the last decade: the Computerized Brain Atlas (CBA) of Thurfjell et al. (1995), the Montreal Brain Atlas of Evans et al. (1994), or the atlas of the International Consortium for Brain Mapping (ICBM) (Mazziotta et al., 2001) are prominent examples. These systems contain probabilistic information on brain anatomy and function
Three-dimensional reconstruction of atlas structures
The printed atlas book of Talairach and Tournoux (1988) consists of three orthogonal stacks of 2D cross-sections through a brain, which are oriented sagittally, coronally and axially. The inter-slice distance is 2–5 mm. Because surgeons often have to consider 3D brain structures and fibre connections between them, we calculated a 3D reconstruction of all objects contained in the atlas at their appropriate location.
For our work we used only the coronal plates. After scanning the 38 plates with a
Design and implementation of an atlas system
In the previous sections we described some methodological basics on 3D reconstruction of the printed Talairach atlas as well as on matching the atlas with MR images of a patient. We used these methods to develop a computerized brain atlas system.
Because all MR scanners supply their images in the DICOM file format, we implemented an import filter for DICOM files. Therefore, the images can be read directly without the necessity of a prior format conversion. A limitation of the import filter is
Three-dimensional reconstruction
The 3D reconstruction algorithm yields accurate and smooth surface models of most of the brain structures contained in the atlas book. However, it is not suited for reconstructing cortex and Brodmann areae, which therefore are not represented three-dimensionally in our atlas system. This is mainly due to the wide gaps between the Talairach plates: there is too much information missing about the complex shape of the cortical convolutions, which can not be compensated by simple interpolation.
As
Summary
We have developed a computerized atlas system based on the Talairach atlas which fulfills the requirements stated in the beginning of this paper (see Section 1). Therefore it not only overcomes principal drawbacks of printed anatomic atlas books, but also the limitations of the Talairach atlas itself: The 3D representation of atlas structures overcomes the two-dimensionality and sparseness of the atlas book, Talairach’s assumption of left/right symmetry of the brain is addressed by the nonrigid
Acknowledgements
We would like to thank the reviewers for their valuable comments on the manuscript.
References (51)
Distance transformations in arbitrary dimensions
Computer Vision Graph. Image Process.
(1984)- et al.
Serial registration of intraoperative MR images of the brain
Med. Image Anal.
(2002) - et al.
Radial basis functions with compact support for elastic registration of medical images
Image Vision Comput.
(2001) - et al.
The use of a computerized brain atlas to support knowledge-based training in radiology
Artif. Intelligence Med.
(1998) - et al.
Simulated brain tumor growth dynamics using a three-dimensional cellular automaton
J. Theor. Biol.
(2000) - et al.
Alignment of magnetic-resonance brain datasets with the stereotactical coordinate system
Med. Image Anal.
(1999) - et al.
Computerized localization of brain structures in single photon emission computed tomography using a proportional anatomical stereotactic atlas
Comput. Med. Imaging Graph.
(1994) - et al.
Neuroinformatics based on VRML
Neuroimage
(1998) - et al.
CBA—an atlas-based software tool used to facilitate the interpretation of neuroimaging data
Comput. Meth. Programs Biomed.
(1995) Error estimates for interpolation by compactly supported radial basis functions of minimal degree
J. Approx. Theory
(1998)
Talairach–Tournoux brain atlas registration using a metal forming principle-based finite element method
Med. Image Anal.
Model creation and deformation for the automatic segmentation of the brain in MR images
IEEE Trans. Biomed. Eng.
Image warping using few anchor points and radial functions
Comput. Graph. Forum
A PC software package to confront multimodality images and a stereotactic atlas in neurosurgery
SPIE Med. Imaging IV: Image Capture and Display
Vergleichende Lokalisationslehre der Großhirnrinde
Spatial normalization of 3D brain images using deformable models
J. Comput. Assist. Tomogr.
Image registration based on boundary mapping
IEEE Trans. Med. Imaging
Three-dimensional Talairach–Tournoux brain atlas
SPIE Med. Imaging 1995: Image Display
Brodmann’s Localisation in the Cerebral Cortex
Cited by (74)
Local statistical deformation models for deformable image registration
2018, NeurocomputingHigh-resolution imaging of the central nervous system: How novel imaging methods combined with navigation strategies will advance patient care
2015, Progress in Brain ResearchCitation Excerpt :In order for readers to compare their patient's brains with the Talairach and Tournoux atlas, the patient's brain MRI scan had to be normalized onto the same coordinate grid as in the atlas, thereby revealing discrepancies between the patient's scan and the atlas and identifying abnormalities (Ganser et al., 2004). MRI atlases, such as the Talairach and Tournoux atlas, are also intended to provide users with high levels of stereotactic and functional information (Ganser et al., 2004). Moreover, due to the popularity of the atlas, the system of coordinate mapping of the brain first described by the atlas has become the most commonly used coordinate brain mapping system (Ganser et al., 2004).
A three-dimensional digital atlas of the dura mater based on human head MRI
2015, Brain ResearchDeformable Registration Algorithm via Non-subsampled Contourlet Transform and Saliency Map
2022, Journal of Shanghai Jiaotong University (Science)