Exploring the brain forest
Section snippets
Global overview of the environment
Modeling brain morphology in one comprehensive framework, from its gross anatomy to its tissue and cellular levels of detail, brings richness to our understanding of brain organization that complements and transcends knowledge derived exclusively from brain atlases and neuron tracing. The human neocortex can be viewed as a 3D shell in physical space with complex geometry. The hierarchical subdivision of the neocortex, from its anatomical lobes to its microscopic neural features, presents a
Framing the virtual environment: The finite element brain
Three-dimensional mesh generation provides a natural structural information framework for the human cerebral cortex. The finite elements of this mesh provide graphical modeling and information management at each hierarchical level, from cellular to tissue to global levels of detail. The finite element model also facilitates global overview and navigation. The resulting finite element model reflects the hierarchical organization of the cerebral cortex and accommodates individual variability.
To
Generating three-dimensional flat maps of the cerebral cortex
There is, however, another way of looking at the finite element decomposition. The cerebral cortex is decomposed through anatomically constrained grid generation into a structured mesh of hexahedral finite elements. The inverse mapping from physical space to configuration space generates a three-dimensional flat map of the cerebral cortex [6], [9]. The top and bottom planar faces of the slab represent 2D flat maps of the outer and inner surfaces of the cerebral cortex, respectively.
Visualization of the finite elements brain
From an initial entry point, the user is able to navigate upward or downward through the hierarchical subdivisions of the neocortex. When the user navigates upward, the initial view is identifiable at the next level by a coloring/highlighting scheme. With 3D flat mapping, navigating around the cortex is a straightforward.
Implanting neurons in finite elements
A human neocortex has been reconstructed from sectional data [2], [6]. Several cytoarchitectural cortical areas of these data have been decomposed into finite elements using grid generation techniques, and rendered. Approximately 6000 finite elements of the scale and complexity shown in Fig. 5 would be needed to model the complete human cerebral cortex.
Finite elements provide a frame for visualizing and modeling neuronal forests. Each neuron in a neuron morphology database, whether a traced
Immersion within its forest of neurons and glial cells
Virtual microscopy, a prototype software system for massively parallel tracing of neurons and mapping their mutual connections, is described in [5]. The system automates tissue scanning, digitizing, feature extraction, and neuron reconstruction. The neuronal environment is then framed in a finite element model of the embedding brain tissue, as described above. We turn now to strategies for visualizing these forests of neurons.
The neuronal forest imagery is sufficiently complex that its
Scale-dependent geometric modeling
The first strategy evokes different geometric models at different ranges (scales) from the viewer (Fig. 7). For extreme close-ups we use implicit function surface models, so the neuronal spines can be visualized [10]. At midrange, we use cylindrical models (typically 7–9 polygons) for dendritic and axonal segments, where a neuron typically may have 200–300 segments. At greater distances we employ wire frame models (with a varying integer thickness of wire) And beyond the range of stereo
Gaze-contingent geometric modeling
The second strategy, guided by parallel ongoing research in our laboratory on gaze-contingent visual communication [8], would deliver to the eye only that minimal level of information that maintains foveal vision while preserving scene preview. This strategy is closely related to the first strategy: scale-dependent geometric modeling. Specifically, a midrange neuron, viewed with the fovea, requires cylindrical modeling; the same neuron viewed in the periphery might be modeled with a wire frame
Summary
In summary, the hexahedral finite element decomposition of the human cortical shell extends facilities that had characterized two-dimensional finite element flat maps into the third dimension, thus providing a natural structural information framework for the human cerebral cortex.
Acknowledgements
This work was supported by Texas Advanced Technology Program grant 999903-124 (McCormick) from the Texas Higher Education Coordinating Board. Virtual reality instrumentation was provided by an NSF Instrumentation Grant for Research in CISE (Volz).
Brent P. Burton is a software engineer at STB Systems in Austin, Texas, performing optimization of 3D graphics drivers for hardware accelerators. He has a B.S. in computer science from Texas A&M University and will complete his M.S. in computer science in December, 1999. His research interests are scientific visualization, and reconstruction techniques as applied to medical image data. Other interests include 3D rendering techniques and programming languages.
References (11)
- D.A. Batte, Finite element decomposition and grid generation for brain modeling and visualization, M.S. Thesis,...
- et al.
Finite element decomposition of human neocortex
- T. Bray, J. Paoli, C. Sperberg-McQueen (Eds.), Extensible Markup Language (XML) 1.0, W3C Recommendation 10-Feb-98,...
- K. Brodmann, Vergleichende Lolkalisationslehre der Grosshirnrinde, Leipzig, J.A. Barth.,...
- B.P. Burton, B.H. McCormick, Virtual microscopy of brain tissue, Neurocomputing, in...
Cited by (0)
Brent P. Burton is a software engineer at STB Systems in Austin, Texas, performing optimization of 3D graphics drivers for hardware accelerators. He has a B.S. in computer science from Texas A&M University and will complete his M.S. in computer science in December, 1999. His research interests are scientific visualization, and reconstruction techniques as applied to medical image data. Other interests include 3D rendering techniques and programming languages.
Travis Seeling Chow received a B.S. in Computer Science from Texas A&M University at College Station in 1996. He worked on his bachelor's degree for three years. He received a M.S. in Computer Science from Texas A&M University in 1998. He has been a Regents Fellow and worked as a graduate teaching assistant and a graduate research assistant for the Computer Science Department. He is currently at Microsoft.
Dr. Duchowski received his B.Sc. ('90) and Ph.D. ('97) degrees in Computer Science from Simon Fraser University, Burnaby, Canada, and Texas A&M University, College Station, TX, respectively. His research and teaching interests include visual attention and perception, eye movements, computer vision, graphics, and virtual environments. He joined the Computer Science faculty at Clemson University in January, 1998 and is currently investigating gaze-contingent virtual reality systems.
Wonryull Koh received a B.S. in Computer Science and Mathematics from the University of Texas at Austin. She is pursuing a M.S. in Computer Science at Texas A&M University. She is interested in computer graphics and scientific visualization. Her thesis research investigates building a structural information framework for the human neocortex.
Dr. McCormick received his B.S. and Ph.D. degrees in Physics from MIT and Harvard University, respectively. He was a professor of Computer Science and Physics at University of Illinois at Urbana-Champaign. At University of Illinois at Chicago and Texas A&M University, he was the department head of the Information Engineering and the Computer Science departments, respectively. He is a professor of Computer Science and the director of the Scientific Visualization Laboratory at Texas A&M University. His research areas include scientific visualization, brain mapping, computer graphics, and neural networks.