Biomedical Engineering Reference
In-Depth Information
data-mining algorithms into one system. At the Erasmus University Medical
Center in Holland [44], researchers have employed a combination of visualiza-
tion and artifi cial intelligence techniques as an important tool for converting
millions of tissue and gene records into straightforward visual information. In
their research, visualization can provide a quick overview of a large amount
of data and help the user narrow down the initial region of interest while the
AI-aided quantitative analysis algorithm further analyzes the data with more
accuracy. Zhou and Liu [45] developed a Java program for visually analyzing
microarray data for gene expression by supporting microarray data visualiza-
tion, quantative assessment, and data mining. Jean Pylouster et al. [46] described
a Web-based tool for gene analysis. This Web tool performs statistical analysis
on gene expression data and identifi es the gene tags that are differentially
expressed and presents plots for the fi nal results. Boyle and his colleagues [47]
developed a software package for exploring embryonic development using
time-lapse confocal imaging and a tree structure to describe the location and
relationship of each nucleus. Their system implements visualization and other
algorithms to extract biological signifi cance out of the data. The combination
of visualization and analytical tools are widely used beyond the above-
mentioned areas. In addition to existing algorithms, techniques, and tools,
current data repositories exist that provide access to abundant community
knowledge. For example, genecards.org provides access to an index database
for gene symbols, and Human Genome Browser Gateway at the University of
California—Santa Cruz [48] provides access to an image-enhanced database
for human genomics.
27.3.2
Technical Approach
The presented visual analytics framework draws from two highly interactive
parallelized display walls termed HIPerWall and HIPerSpace, fi rst commis-
sioned in 2005 and 2007, respectively, capitalizing on custom-developed mid-
dleware called CGLX. The visual analytics tools in turn use both of these to
provide an intuitive and collaborative digital workspace.
27.3.2.1 Hardware Confi guration
HIPerWall utilizes fi fty 30 - in. displays
with a resolution of 2560
×
1600 pixels each, confi gured in a 10
×
5 (width - by -
length) layout, resulting in a combined resolution of 25,600
8000 pixels
(204,800,000 pixels total). HIPerSpace in turn uses 70 display tiles in a 14
×
×
5
layout, resulting in 35,840
8000 pixels resolution (286,720,000 pixels total).
The computing and rendering cluster of HIPerWall is based on 25 Power Mac
G5 running OS-X, with a 2.7-GHz IBM PowerPC processor, 2 GB of RAM,
with dual, dual-core processors, and NVIDIA Quadro FX 4500 graphics.
Each one of these HIPerWall nodes drives two displays and is interconnected
in a dedicated gigabit subnet. Data access is provided via a dedicated, nfs
mounted storage node (HIPerStore), while a stand-alone control node serves
as the front end. Similarly, HIPerSpace utilizes 18 machines running Linux,
×
Search WWH ::
Custom Search