Next: References Previous: Title
Page Contents: Conference page
The Distributed Visible Human Explorer
Introduction
The Visible Human Explorer from
Flashback Imaging is an
innovative approach to fast and interactive access to the full image dataset of the Visible Man
and Visible Woman
(RSNA'96
InfoRad). It incorporates reconstructed coronal and sagittal views and is targeted for
economical personal computer platforms. The Visible Human Explorer is available in many
forms and sizes, ranging from a simple
CD-based
system to a full-scale 30-gigabyte version. In the last few years, the stand-alone
exhibit version with
30,000 images derived from the Visible Human Project has been installed in a number of large
science centres and museums in Canada and the USA. The visitors to these exhibits freely roam
through the male or female body, interactively browsing 30,000 high-resolution full color
anatomical images.
While the Visible Human Explorer exhibits have allowed millions of children and adults to learn
a great deal about the Visible Human Project, the lack of high quality segmented and labeled
datasets, in particular for the Visible Woman, has been an obstacle to further educate the public.
One way to speed up the segmentation and labeling of the Visible Woman dataset is to come up
with a system that allows large-scale collaboration among anatomists and related personnel who
will share the workload of this enormous task. Starting with a simple Local Area Network with
the goal to expand to the World Wide Web, Flashback Imaging is attempting to derive a
distributed version of the Visible Human Explorer that helps achieve this objective.
Methods
While the networked version allows multiple accesses to the same dataset, the objective of the
Distributed Visible Human Explorer is not just data sharing. The plan is to form a work-group,
collaborative scheme for labeling and segmenting the data, in particular the complete Visible
Woman dataset. The approach is to provide a customized software toolkit that allows widely
distributed experts to quickly and easily annotate individual slices while almost immediately
verifying annotations and changes with their colleagues over a wide area network. With the
scalable client/server architecture, many more specialists can join the team and be part of an
integrated process at any time. The customized interface based on the Visible Human Explorer
engine increases efficiency and quality control. Using a combination of local and remote storage
devices, the same collaboration can be extended to the entire network of the Internet. This
ultimately allows greater international collaboration and quicker access and distribution of the
information content inherent in the Visible Human dataset.
Results
In the first phase of the project, we placed the standard 30,000 image set used in exhibition on an
NT server connected to a local area network of a few thousand nodes (mixed between 10BaseT
and 100BaseT). The network-enabled Visible Human Explorer software is also installed on the
same PC. From any other PC over the network, access to the complete Visible Human database
is done by simply pointing and clicking at the executable file located on the server. No
installation is needed from any client PC. This setup forms the most basic version of the
Distributed Visible Human Explorer.
The Visible Human Explorer based on this client/server model allows the entire digital library of
the Visible Human data, including the reconstructed coronal and sagittal data, to be stored at a
single central location. The software permits interactive access to the full image datasets from
remote locations networked to the central site. It allows sharing of the datasets among multiple
users while offering similar speed and interactivity as the single user, stand-alone system.
Recently, this distributed version has been installed in the
Samsung Medical Center
in Seoul and the
SungKyunKwan
University in Suwon, Korea. The installation was done on the regular fast Ethernet
networks with a few thousand nodes each. Initial results indicate that the browsing speed of the
images ranges from excellent to reasonably good during low traffic time, and less than desireable
when the traffic is heavy. In this early model, images are moving across the network in
non-compressed mode, which places heavy load on the system. We are studying several
compression technologies that allow image size reduction between 10 times and 30 times.
Current tests indicate that speed improvement is between 2 times (in low-traffic time) to 10 times
(in high-traffic time).
Conclusions
As we are working toward our goal of completing the work-group system, we realize that the
same system should also serves as a model for installation on faster networks such as the Gigabit
network and the Next Generation Internet. High quality images could be moved quickly across
the Internet which would further enhance the experience with interactivity.
The core software engine
used in the Visible Human Explorer is suitable for any similar tasks and can be used as an
interactive search engine for other large databases
(e.g., remote sensing)
with an appropriate user interface. Since the capability of the engine is virtually independent of
the database size, this approach is ready for the future challenge of increasingly large data
volumes.
The technologies are designed to solve the basic problem of information accessing: "too much
information, too little time". Future web-based applications based on the same technologies will
further the goal and has the potential of changing the way that most users search information over
the web.
Next: References Previous: Title
Page Contents: Conference page
Office of High Performance Computing and
Communications
Lister Hill National Center for Biomedical
Communications
U.S. National Library of
Medicine, 8600 Rockville Pike, Bethesda, MD
20894
National Institutes of Health
Department of Health & Human Services
Copyright and Privacy Policy
Last updated: 2 July 2001
|
