Java Internet Viewer: a WWW Tool for
Remote 3D Medical Image Data Visualization and Comparison

CHRIS A. COCOSCO and ALAN C. EVANS

McConnell Brain Imaging Centre, Montréal Neurological Institute, McGill University, Montréal, Canada

email: c.cocosco@ieee.org

http://www.bic.mni.mcgill.ca/users/crisco/jiv/

Introduction:

There is a growing need in the research and clinical medical imaging community for Internet-capable tools that facilitate remote data dissemination and interaction. 3-dimensional (3D) medical imaging datasets typically require special-purpose, non-portable, software to be installed and maintained on each workstation.

We developed ``JIV'': a powerful, robust, portable, extensible, and open-source Java (v 1.1) application (applet) for visualization and side-by-side comparison of multiple 3D image datasets. JIV is designed to work through the WWW.

A convenient way to visualize 3D medical imaging datasets is by three orthogonal 2D slices through the same location in the volume. When several image volumes are to be compared, it is desirable to visualize their slices side-by-side, all at the same position in the volume (fig. 1).

$\psfig{figure=figs/jiv_many2.eps,width=0.99\linewidth}$

$\textstyle \parbox{0.9\linewidth}{ \figurenum Screenshot of the JIV interfac... ...lumn from the left). Some of the slices are zoomed-in, by different amounts. }$

JIV Features:

$\bullet$: requires no installation and no maintenance on the workstation, but only a common Web browser.
$\bullet$: highly portable: the vast majority of computers come with Web browsers, which support Java 1.1. JIV proved to be robust on a variety of computer platforms: Linux-i386, SGI IRIX, MacOS, and various MS Windows (Win32) versions.
$\bullet$: can cope with slow network links.
$\bullet$: uses a simple and platform-independent image data format, and accepts arbitrary volume samplings.
$\bullet$: displays gray-level (intensity) image data with a user-controlled color-mapping; also provides merged (blended) views of two datasets.
$\bullet$: an efficient mouse-based user input scheme provides fast roaming through the volume, and continuously adjustable zoom and pan.

Design & implementation:

In order to provide remote data access with a level of performance comparable to traditional stand-alone workstation applications, the issue of file loading (``I/O''), which is significantly slower, needs to be addressed. In practice, the performance of regular Internet connections is unpredictable; moreover, the transfer rate can vary by up to three orders of magnitude (1000x) among different types of network connections. With respect to how and when to download the 3D image data, the following three operation modes are supported by JIV:

All up-front: all of the data is downloaded and stored in client's memory before the user can view and interact with any of it. This guarantees the best interactive response of the viewer; however, the user has to wait for all the data to download before the JIV interface becomes available, which can be impractical.
On demand: download slice image data only when and if the user wants to view that particular slice. This minimizes the data downloads (and the amount of memory required by the applet), but the interactive response time is highly dependent on the server and on the network speed.
Hybrid (background download): first download only the slices required by the initial cursor positions; then continue downloading all of the data in a background thread (as in (1)); if the user requests slice images which are not already downloaded, they will be downloaded with priority (as in (2)).

The hybrid mode (see fig. 2) is the best compromise for most practical situations, and can be seen as a predictive download approach. JIV could be configured to first transfer the slices with high probability of being visited by the user (this probability is application and data dependent); such a technique will increase the likelihood of the image data to be already available when the user moves the cursor to a new location.

$\psfig{figure=figs/timeline.eps,width=0.99\linewidth}$

$\textstyle \parbox{0.9\linewidth}{ \figurenum The hybrid download functional... ... sampling brain MRI, compressed with \texttt{gnuzip} (lossless compression). }$

$\psfig{figure=figs/jiv_merge_2.eps,width=0.95\linewidth}$

$\psfig{figure=figs/jiv_merge_1.eps,width=0.95\linewidth}$

$\textstyle \parbox{0.9\linewidth}{ \figurenum Sample JIV application: evalua... ... sliders (window/level) can be used to restrict the visible tissue classes. }$

Results & Conclusion:

An online demo as well as the software are available at the Web address in the title.

JIV is a convenient and platform-independent software for the remote visualization of 3D medical image data; for example, it can be used in remote data processing - when data goes to a central, well-equipped, site for image processing and storage. This software proved useful (both remotely, as an applet, and as a stand-alone application) for efficient simultaneous visualization and comparison of many 3D image datasets, such as evaluating the performance of registration or segmentation methods (fig. 3). It can also be used as a PACS viewer in a radiology environment.

Acknowledgments: Peter Neelin, Alex Zijdenbos, Steve Robbins, Louis Collins, John Sled.

About this document ...

This document was generated using the LaTeX2HTML translator Version 2K.1beta (1.49)

The command line arguments were:
latex2html -split 0 -mkdir -dir /data/nil/crisco/www/jiv/miccai01_poster -title 'JIV MICCAI 2001 poster' p3.tex

The translation was initiated by Chris COCOSCO on 2001-10-11

Chris COCOSCO 2001-10-11