Presenter: 

Lars Linsen
Center for Image Processing and Integrated Computing (CIPIC)
Department of Computer Science 
University of California, Davis 
2144 Academic Surge
Davis, CA 95616
http://graphics.cs.ucdavis.edu
llinsen@ucdavis.edu

Authors:

Lars Linsen*, Alfred R. Fuller*, Oliver Kreylos*, Giorgio Scorzelli*,
Fabien Vivodtzev*, Patrick C. Yau*, Bernd Hamann*, Kenneth I. Joy*, 
Bruno A. Olshausen^, and Edward G. Jones^
* Center for Image Processing and Integrated Computing (CIPIC), 
  University of California, Davis
^ Center for Neuroscience, University of California, Davis

Title: 

Visual Exploration of High-resolution Neuroscientific Data

Abstract:

We have developed scalable, interactive rendering techniques for 
large-scale volumetric data sets and its application to 3D brain 
atlases and other related neuroscientific data. 
Several multiresolution rendering techniques were emphasized and 
developed, including the functionality of brain-mapping techniques,
where annotations from a source brain are mapped to a target brain.

A progressive arbitrary slicing tool has been developed,
which is capable of visualizing cutting planes through large-scale
volumetric data sets with interactive frame rates. It makes use of
a hierarchical indexing storage scheme (hierarchical space-filling
curves) to speed up the data loading / data exploration. Further 
speed-ups are achieved by using parallel algorithms / distributed 
computing on a Linux cluster.
The data is distributed on local hard disks of several "data 
servers". A client sends requests to all servers, which process 
the incoming data simultaneously and deliver the relevant data 
informations. The clients are in charge of user interaction and of
rendering of the arbitrary cutting planes. 
The slicing tool has been applied to many large-scale brain data 
sets.

For a more volumetric representation and visualization of 
neuroscientific data, We have developed a volume renderer that 
works for and is based on the AMR data structure. AMR provides a 
data set on several rectilinear structured grids of different 
levels of resolution. These grids can be overlapping and are not
nested. The idea is to split up the AMR data set into 
non-overlapping grid patches that can be visualized using standard
direct volume rendering methods. A remotely working approach on AMR
data sets can be used for massive-parallel computations. The AMR
volume renderer has been applied for the visualization of ganglion
retina cells.

A high-level approach to describe the characteristics of a surface
is to segment it into regions with the same type of curvature 
behavior and construct an abstract representation, represented, for
example, by a (topology) graph. We have developed a surface 
segmentation method based on discrete mean and Gaussian curvature
estimates. The surfaces were obtained from three-dimensional MRI
imaging data sets by isosurface extraction after data pre-smoothing
and post-processing the isosurfaces by a surface-growing algorithm. 
We generate a hierarchical multiresolution representation of the
isosurface. Segmentation and graph generation algorithms can be 
performed at various levels of detail. At a coarse level of detail,
our algorithm detects the main features of the surface. This 
low-resolution description is used to determine constraints for
the segmentation and graph generation at the higher resolutions. 
We have applied our methods to MRI data sets of human brains. The
hierarchical segmentation framework can be used for brain-mapping 
purposes.