Biomedical Engineering Reference
In-Depth Information
27.1
MOTIVATION: THE LARGER CHALLENGE
The annual amount of information the world produces has been at a continu-
ously increasing pace over the past decade. Varian and Lyman found that in
the early part of the 2000s the world produced between one and two exabytes
(one billion gigabytes, or 10
18
bytes) of unique information per year [1]. A
white paper by the International Data Corporation (IDC), sponsored by EMC,
found that this number had jumped to 161 exabytes by 2006 [2]. This paper
projected that by the year 2010 this number could jump to 988 exabytes per
year. This information overload has left many researchers scrambling to fi nd
ways to analyze this kind of massive amount of data. While well-defi ned tasks
can be delegated to computer processing, oftentimes, researchers may desire
to approach data analysis from a more exploratory perspective. In this way,
researchers may attempt to fi nd characteristics and/or patterns of interest in
the available information. Similar to the overall data explosion across the
fi elds of science and engineering, medicine and biomedical engineering have
seen their own data avalanche through the introduction of new imaging, sam-
pling, and modeling techniques. As a result, domain experts are pressed to
create new methodologies, techniques, and tools to manage and most of all
harness this wealth of information.
The fi eld of visual analytics attempts to approach these problems through
human - centric visualization. Thomas defi nes visual analytics as the “science
of analytical reasoning facilitated by interactive visual interfaces” [3, p. 4].
The goal of visual analytics is to present data in such a way that the human
mind is able to effi ciently process information, combining the benefi ts of
machine and human analysis. Thomas proposes that visual analytics has the
ability to “detect the expected and discover the unexpected” [3]. The goal of
this paradigm is to allow researchers to visually explore data sets without
explicitly knowing what they are looking for initially. For general-purpose
pattern recognition, the human brain can outperform machine-based algo-
rithms [4]. It only takes the human brain a little over a tenth of a second to
identify and classify an object in a complicated environment [5]. Furthermore,
the human brain can fi nd patterns and differences even when the differences
seen in objects are not easily quantifi able, that is, the symbol grounding
problem [6] .
The challenge for visual analytics is that data must be organized and pre-
sented in a meaningful way to be effective for analysts. For example, the visual
analytics paradigms useful for large image collection will most likely be dif-
ferent from those for detecting intruders on a network. Visual analytics tech-
niques need to be customized for the data being analyzed as well the users of
the system. This is to say, there is no “one size fi ts all” solution. As such, this
chapter focuses on a high-level overview with focus on large-scale multidimen-
sional data analysis in environments suitable for team-based visual analytics.
Commonly encountered data in biomedical research consist of multidimen-
sional, layered two - dimensional data and three - dimensional volumetric data
Search WWH ::
Custom Search