A computer scientist at Washington University in St. Louis has adapted the hottest programming language of the Internet and applied it to a burgeoning revolution in medicine - teleradiology.
Douglas C. Schmidt, Ph.D., associate professor of computer science at Washington University, used Java programming language and the popular Web browsers Netscape and the Internet Explorer to effectively perform medical imaging tasks, moving medicine more deeply into the virtual world.
Electronic radiology refers to the conversion from a film-based technology to digital image acquisition, transmission, storage and display on cathode-ray tube devices. Teleradiology involves the transmission and examination of images and X-rays by specialists either buildings or miles apart. It is practiced in varying degrees at hospitals nationwide, including the Washington University School of Medicine in St. Louis and its multihospital affiliate, BJC Health System.
Once realized on a large scale, teleradiology will eliminate the costly use of film for imaging and make the entire process digital and electronic. A major attraction of teleradiology is the possible re-engineering of radiology practice. Quick and easy access to medical images promises to save physician time, as well as support the concept of "centers of expertise." This should translate into dramatic cost savings.
Teleradiology opens the possibility of remote diagnostic conferencing, and, eventually, even distance surgery. A physician in a Los Angeles hotel could access the Internet via a laptop computer to examine instantaneously a patient's MRI or CT image taken in St. Louis and consult with doctors anywhere in the world.
At Washington University School of Medicine and BJC Health System, an ambitious project is under way to provide physicians access to an integrated clinical record of medical images and data. In addition to Washington University and BJC, Project Spectrum is co-sponsored by Kodak Co., IBM, Southwestern Bell and Motorola Co. James Blaine, D.Sc., professor of radiology at the School of Medicine, serves as the project director for the imaging task. Blaine says images now are available for distribution from a central server at the medical center in St. Louis to physicians' personal computers at home using special software over an "intranet," an ISDN network available through Southwestern Bell.
Washington University's Mallinckrodt Institute of Radiology also uses ISDN to support "off-hours teleradiology" to expedite interpretation of radiology studies at BJC satellite hospitals. Several BJC hospitals are linked to the medical center hospitals via a high-speed network enabling the transmission of CT and MRI images.
MedJava is the name of Schmidt's electronic medical imaging system (EMIS). He and his colleagues developed the system and then tested its performance in transmitting 8-bit images through eight different imaging filters a total of 10 times, keeping the average of the trials. He benchmarked MedJava against a popular image-viewing application called xv, which runs on the common languages C and C++. Both applications were tested on an ultra-fast high speed switching network called asynchronous transfer mode.
As Schmidt expected, xv had the edge over MedJava in speed and image quality, but MedJava performed far better than expected with good, distinctive images and decent speed. And MedJava has the advantage of flexibility and portability over xv. It did show one limitation. Today's browsers are not yet capable of rendering and displaying large images using MedJava because of memory constraints. However, Schmidt contends, browsers soon will be modified to handle that task.
"We were actually surprised that the Java performance was so good for these very computing-intensive image processing applications," says Schmidt. "We've shown that MedJava is particularly good for teleradiology, an application where flexibility and portability become more important than raw performance. MedJava can do very remote medical imaging processing without the expensive installation of special software in machines such as MRI or CT scanners because all of the software needed can be downloaded into the browser on demand. This makes MedJava more flexible because all of those software upgrades are accomplished with the browser."
Schmidt compares MedJava's service quality to that of a very popular device millions of Americans now own - the cellular phone.
"People don't expect as much quality from their cell phones compared with their office or home phones," he says. "But the fact that they can make a call while driving or else from some other inaccessible location more than makes up for the lesser quality. The same is true for medical imaging technology. In the bigger picture, our adaptation of Java language is an enabling technique that will facilitate the growth of industries that can increasingly leverage the accessibility and wealth of the World Wide Web."
Maturing technology
Schmidt's paper, "The Design and Performance of MedJava," appears in the December 1998 issue of "IEE/BCS Distributed Systems Engineering Journal." An earlier presentation of the paper at the 4th USENIX Conference on Object-Oriented Technologies and Systems won Schmidt a "Best Paper" citation at the conference.
Medical imaging giant Siemens Medical Engineering, in Erlangen, Germany, supported part of the research and is now building their version of MedJava for an array of medical imaging devices, ranging from CT and MRI machines to angiography and ultrasound technologies. Other companies may find the application beneficial for mobile computing, electronic commerce or any sort of technology involving imaging.
"More and more hospitals are moving toward 'filmless' radiology," notes Blaine. "When you go filmless, the question becomes: 'How do you get the images to the attending physician'" Browser technology plays a key role in making this happen."
Schmidt recognized that newer generations of Java, with improved interfaces between Web browsers and Java Virtual Machines, which execute a host of Java re-usable tools called classes, would be able to handle the task with the aid of compilers called "just-in-time" compilers (JIT). These translate Java code into code other machines might be using right before the machine performs the application, allowing quicker rendering and image-processing.
"Large health care systems want to move toward a digital world where everything is stored on a data base and accessible on a major network and moved at high speeds all over the place," Schmidt says. "We expect that Java technology will continue to mature, and with that maturity people will look toward Java for more challenging applications. Medical imaging systems will be just one small part of a larger technological fabric."