Skip to Content Sun Java Solaris Communities My SDN Account Join SDN

SDN Home > Products & Technologies > Java Technology > Reference > Technical Articles and Tips > Feature Stories About Java Technology > Feature Stories About Java Technology - 2001 >

Article

In Living Color

Print-friendly Version

by Steve Meloan

January 29, 2001 -- With Java technology now finding its way into nearly every industry and functional niche, it comes as no surprise to discover Java applets, applications, servlets, and more, being used in the exploding field of medical imaging. In October 2000, java.sun.com profiled Emageon, Inc., a Java technology-based medical imaging software start-up. Now, the National Library of Medicine (NLM), a division of the U.S. government's National Institutes of Health (NIH), is in the beta test phase of AnatLine, a prototype system consisting of an anatomical image database, and an online browser for accessing the data.

AnatLine was developed by a five-member team at the NLM's Lister Hill National Center for Biomedical Communications. "The NLM is the largest repository and distributor of medical and health information in the U.S.," explains Earl Henderson, deputy director of the Lister Hill Center. "Its databases are accessed by national and international health care providers, as well as by academics and research professionals. The Lister Hill Center, meanwhile, is a research division of the NLM, with programs targeted to advancing the collection, storage, and distribution of biomedical information."

AnatLine 3D Imaging Terms

Voxel: The smallest box-shaped element of a three-dimensional image.

Cross-Section File: A single plane, axial view of the body.

Segmentation: The process of isolating certain objects or structures from surrounding data, or of partitioning an image into meaningful areas. Also used when creating 3D objects from a stack of 2D images.

Segmented Structure: A subcomponent extracted from a larger composite image using a segmentation process. Also used to refer to an extracted anatomical structure, such as the heart.

Segmented Bytemask File: A bitmap file structure used to store collections of related anatomical segmented volume structures in an encoded format, such that each structure is defined and retrievable.

Volume of Interest File: Voxel data of individual anatomical structures, such as heart, lung, etc. These files contain color regions, and segmented masks for individual structures.

Rendered Images: Often refers to the process of adding realism to a computer-generated image by adding three-dimensional qualities, such as shadows and variations in color and shade. But AnatLine's rendered image file format can store all types of media productions, including audiovisuals and animation.

Hardware/Software Requirements Necessary to Access the AnatLine System

The AnatLine system is intended for medical and biological users with workstations connected to high-speed networks. It is also recommended that users have either Unix^® or Microsoft Windows (NT, 98) systems, with 512MB of RAM, and using Netscape Navigator 4.75, Internet Explorer 5.0 or higher. (Note: The Macintosh platform will be supported in early 2001.)

Since downloaded files can range from 1MB to 1GB in size, substantial disk space must also be available. And because of the large file sizes involved, a T1 (or higher) network connection is recommended. Until January 2000, only registered/approved beta users were given password access to the full facilities of the AnatLine system. But the system has now been opened to full public access. The beta phase will complete later this year.

The Visible Human Project

In 1986, the NLM foresaw an era in which its database services would be complemented by libraries of digital images, distributed over high-speed networks, and stored on high-capacity computer media. In 1989, a planning panel convened and recommended that the NLM should embark on a prototype project of building "a digital image library of volumetric data, representing a complete, normal adult male and female." This library was to include imaging from computerized tomography (CT scans), and magnetic resonance imaging (MRI scans) of cadavers. In short, the panel recommended that the NLM begin to build and disseminate medical imaging libraries in the same way that it acquired, indexed, and provided access to biomedical literature.

This image data acquisition project was awarded to a team at the University of Colorado at Denver. The complete male data set, 15GB in size, was made available in November 1994. The female data set (adjusted along the Z-axis for greater resolution) was released in November 1995, at a size of 40GB.

Once the data was acquired, the next step was to devise a means of searching, distributing, and interpreting the information. With these goals in mind, the NLM has convened a Visible Human Project Users' Conference every two years since 1996. There, licensees have demonstrated products, and discussed models and algorithms based on the image data sets.

AnatLine and Java Technology

The second phase of the Visible Human Project was soon well under way. A team of image processors at Engineering Animation, Inc. performed the alignment and segmentation of the male thorax cross-sectional data.

But it was in the design of the image database that the real technical challenges began. "Our initial problem in designing the database was related to data structures," explains Henderson. "Normally, when one designs a database, it is targeted to explore some form of existing data. Of course we knew the basic form of the data, that it would be made up of cross-sectional images. But the structure of the voxel data for 3D volumes remained to be specified."

Converting anatomical images into functional data types presented unique problems. "Anatomical images are not joinable in the same way as other data types," adds Henderson. "Within a structure like the heart, you have male, female, age groups, and pathology conditions. Then there are the various media types to consider -- CT, MRI, color, etc. And that does not even begin to address any future data types."

The need for creative data structures, an interface to complex data, and image processing functions, soon led the development team to conclude that an object-oriented design approach was the only way to go. "We needed an object-oriented database with self-contained class structures," explains Henderson.

The first working prototype of the database was implemented using C++. But once the team began exploring the actual delivery of image content to their user base, it became clear that they also needed cross-platform compatibility. Around this same time, Jim Seamans, senior software engineer for the Lister Hill development team, began exploring Java technology. "As the design team considered several options for designing the data delivery system, Seamans provided the strongest arguments in favor of using Java technology, which at that time was emerging as a viable object-oriented language," explains Henderson. "And most importantly, coding in the Java programming language offered the solution to our need for cross-platform compatibility."

The project was soon off and running using Java technology. "Over a period of three years," says Henderson, "following the initial prototype of the system, we have replaced all of our C++ code with Java technology, and are coding all new modules in the Java programming language. We are currently using Java applets, applications, servlets, the Java 3D API, and the Java Advanced Imaging API in our high-resolution display module."

Up and Running

AnatLine was developed to provide a client-server interface to access gross anatomy images of the human body. Its data structures are based on anatomical and spatial relationships of gross anatomy components and physiological systems. Actual image records are stored in the NLM's prototype .vhi (Visible Human Image) file format. During this beta phase, AnatLine is being used to store only the male thorax data sets of the Visible Human Project.

The current implementation of AnatLine consists of:

An anatomical image database, with applet-based query and retrieval capabilities.
A download capability for delivering single and multiple images.
Java application-based utilities to both view and parse the high-resolution images.
A Java applet-based browser that is offered as a prototype facility to navigate the database via a visual/anatomical interface. This browser facility provides access to low-resolution rendered images of anatomical structures.

Under the Hood

Behind the scenes, servlets support the RMI interface between the online server (currently, a Sun Enterprise 5500 server) and the actual database. "Users do not have direct access to the database," explains Henderson. "The server controls the query, retrieval, and download process."


Image 1	Image 2

Being applet based, the query and retrieval engine offers a spell-assist feature for complex anatomical/physiological terms (see image 1). A pop-up window then displays those anatomical terms that contain the search entry (image 2). Successful queries return results containing embedded hyperlinks to retrievable image types (see image 3), and links to anatomical objects related to the queried structure -- that is, anatomical parent, substructures, physiological system, and body region (see image 4).


Image 3	Image 4

There are four types of image files made available for retrieval by AnatLine -- labeled cross section, volume of interest, segmented mask, and rendered image. Each file type stores a complex format of data maps, data tables, and image bitmaps. This image data can be accessed via AnatLine's Java application-based user utilities, consisting of a high-resolution display facility (VHDisplay), and an image parser (VHParser) used to extract specific segmented images from the volume image records.


Image 5	Image 6	Image 7

Meanwhile, the applet-based anatomical browser offers a prototype visual navigational tool into the low-resolution rendered images. The human male body acts as the top image node, branching down into the thorax (the only choice currently available). From there, additional pop-up applets provide visual navigation into rendered images of the thoracic wall (see image 5, 6, 7), the thoracic cavity, the blood vessels of the thorax (see image 8, 9, 10), and the spine.


Image 8	Image 9	Image 10

Constructing Medical Imaging Libraries

The long-term goal of the Visible Human Project and AnatLine is to act as a test bed for the construction of medical imaging libraries that can be accessed through high-bandwidth computer networks. But with the wild proliferation of new technologies ever around the corner, such a goal is a perpetual work-in-progress. "Future incarnations of the project, utilizing the Next Generation Internet, will include such facilities as real-time 2D and 3D visualizations under haptic (touch-sensitive) control, as well as virtual reality applications," reports Henderson.

Beyond these enhancements, at a functional level, a key challenge is to effectively link image-based structural-anatomical data to text-based functional-physiological information -- presenting them as a single, unified source of medical information. But here again, the scalable, cross-platform, object-oriented nature of Java technology should serve the project well, effectively rendering it future-proof.