There has been much talk about the use of Web technology in medical imaging, specifically its application in picture archiving and communication systems (PACS) and teleradiology. This article will define Web technology and its associated ter-minology, and discuss its relevance to medical imaging.

What is the "Web"?

The term "Web" is used to mean many things. Usually, it is understood to mean a community of computers and users communicating over the Internet, making use of Web browsers, e-mail utilities, and other technolo-gies. At the highest level, where most of us function, there is no distinction made between "the Web" and "the Internet" or the technologies that make it all possible. It is this lack of distinction that causes significant confusion in technical discussions of PACS and teleradiology, as well as confusion among potential purchasers of these systems.

Web technology is not a single technology, it is a multiplicity of technologies. Table 1 provides a (partial) list of the technologies usually consid-ered Web technology.

Web technology in PACS and teleradiology applications

Although the technologies listed in Table 1 (and others) are frequently thought of as Web technologies or an integral part of the Web, their useful-ness in PACS and teleradiology appli-cations must be examined individually, not as a whole. Some of these technologies are highly useful in PACS and teleradiology systems, while others are considerably less so.

Fundamental technologies, like TCP/IP and the Internet, are very useful for PACS and teleradiology deployment. TCP/IP has become the de-facto, low-level standard protocol for implementation of the DICOM standard. The Internet provides a low cost communication infrastructure, which can be useful in teleradiology, remote viewing, and report distribu-tion applications. The Internet does have its drawbacks, however: unreliable communication bandwidth and significant security problems.

Other than these two technologies, the others mentioned in Table 1 have much more limited utility in PACS and teleradiology. XML shows some promise for the future, but is of lim-ited use today, other than in proprietary applications. The DICOM, HL-7, and other standards bodies are studying ways in which XML could be used in a "standard fashion" to address medical informatics problems. Such standardization is years away, however. This standardization is also of questionable utility, as DICOM today addresses many, if not most, of the issues that could be addressed by XML.

JAVA is really a programming tool. As such, it should be of no interest to a PACS or teleradiology user or pur-chaser, other than if the use of this technology results in a more reliable, more maintainable, higher performing, and/or lower cost system. JAVA usage today, however, does not guarantee any of these.

Web browser/server, hypertext trans-fer protocol (HTTP), hypertext markup language (HTML), and file transfer protocol (FTP) are technologies that can be used to address PACS and teleradiology problems. These technologies however, cannot address all problems or any individual problem in an optimal fashion. As an example, FTP is a far less useful protocol for medical imaging applications than DICOM. FTP can be used to transfer medical images from one computer to another, but in a far less elegant and ultimately useful fash-ion than allowed for in the DICOM standard. HTTP and HTML technolo-gies are only useful in the context of Web browser and Web server applica-tions. The application most often associated with Web browser/server technology is teleradiology.

Web server-based teleradiology

In its simplest form, a Web server-based teleradiology system works as follows:

1) Images are received from imag-ing systems in the radiology department (generally by means of DICOM).

2) The received images are con-verted into a standard image format, such as JPEG or TIFF, and stored on the server. The demographic data associated with an image is stripped out and placed in its own text file. The image file and the text file are associated with each other by means of a computer directory structure, or database.

3) A remote user connects to the server, over the Internet, using a general purpose Web browser without any additional software.

Table 1. Definitions of Web technology terms

Internet A publicly owned, network-based communication infrastructure that grew out of DARPA (Defense Advanced Research Project Agency) research in the 1970s and 1980s. The Internet relies on the TCP/IP communication protocol for its addressing and routing capabilities.

Web browser A software program that resides on a computer that allows a user to view pages of information supplied by Web Servers.

Web server A computer and software program that stores pages of information and presents that information on request of a remote user using a Web browser.

TCP/IP A low level communication protocol that defines the addressing, routing, and message transfer protocols used by higher-level services. It can be thought of as an electronic envelope with address that is delivered by an electronic postal service. TCP/IP defines the envelope and the rules used by the elec-tronic postal service to deliver the envelope.

HTTP Hypertext transfer protocol: An application-level communication protocol or service used for communication between Web browsers and Web servers over a TCP/IP network. It can be thought of as a specially formatted form letter, placed in an electronic envelope.

FTP File transfer protocol: An application-level communication protocol or service used for transferring data files from one computer to another over a TCP/IP network. It can be thought of as a specially formatted form letter, placed in an electronic envelope.

HTML Hypertext markup language: A language used to define the formatting of Web. It is very similar in concept to the "language" used to format an electronic document generated by word processing or desktop publishing software. HTML is non-proprietary and less full-featured than most other page format description languages.

JAVA A high-level programming language, originally developed by Sun Microsystems. The goal of this language is to pro-vide an object-oriented programming language that is totally hardware independent.

XML Extensible markup language: This language similar to HTML, but with more ambitious goals. XML is intended as a document formatting language, capable of defining and managing a broader set of document "objects," than just text, graphics, and images. In the future, XML may have the ability to address the same types of problems (data representation and usage) as the DICOM standard currently does.

4) The user is presented with a list (directory) of viewable images by means of a Web page formatted in HTML and transferred to the remote user over the Internet, using HTTP.

5) Upon selecting an image to view, the user causes the Web server to build a Web page that contains the requested image and the information from the associated demographic text file. This Web page is transferred to the remote user over the Internet, using HTTP.

This simple Web server-based tel-eradiology model has a number of dis-advantages. It is not possible to window or level the retrieved image, as such functionality is not part of a general purpose Web browser. In order to accomplish this function, the Web server would have to constantly send a new Web page with the same image, windowed and leveled differ-ently by the Web server. It may also be that the image was truncated to an 8- bit image when converted to JPEG or TIFF, making window/level meaningless. Specialty viewing functions (like cine, stack, dynamic multi-image format definition, annotation, zoom, etc.) are not available (at least, not easily) inside this simple model, for the same reason that it is not possible to window and level in a useful fashion. The only justification for such a simple model is cost. The required Web browser application is inexpensive and generally available. This model, however, is unsuitable for any appliation other than the transfer of an image with a text report to referring physicians for record keeping and/or patient consultation purposes.

It is possible to overcome the limitations inherent in this model, by cre-ating a specialty Web browser or tailoring a general purpose Web brow-ser by means of applets and/or plug-ins, which can be thought of as application accessories. Both of these solutions require engineering development on the part of a vendor, how-ever, which no longer makes the viewing application "free" (even if the vendor gives away the applet or plug-in). It also requires significant engi-neering effort to make these custom browsers perform well.

DICOM-based pull teleradiology

A DICOM based "pull" teleradiol-ogy system, in its simplest form, works as follows: 1) Images are received from imag-ing systems in the radiology depart-ment (generally by means of DICOM).

2) The received images are stored as DICOM files on the teleradiology server.

3) A remote user connects to the server, over the Internet, using specialty teleradiology application soft-ware.

4) Making use of DICOM query/ retrieve protocols, images of interest are transferred across the Internet to the remote user.

5) The remote user now makes use of the specialty teleradiology application software to view and manipulate the images in a fashion similar to how these operations would be performed on a dedicated radiology viewing station in the radiology department.

DICOM-based push teleradiology

Most simply, a DICOM-based push teleradiology system works as follows:

1) Images are received from imag-ing systems in the radiology department (generally by means of DICOM).

2) The received images are stored as DICOM files on the teleradiology server.

3) A routing agent (software) in the teleradiology server determines to whom the images are to be distributed and attempts to forward the images to that user.

4) A remote user connects to the server, over the Internet, using a static IP address, by means of specialty teleradiology application software.

5) The teleradiology server for-wards the images to the remote user.

6) The remote user now makes use of specialty teleradiology application software to view and manipulate the images in a fashion similar to how these operations would be performed on a dedicated radiology viewing station in the radiology department.

The DICOM-based teleradiology systems described have the ability to accomplish the same tasks as a Web server-based teleradiology system. All three types of systems are capable of making use of the Internet. So how does one choose between them?

Web technology low cost myth

It is an article of faith, today, that using the Web results in lower cost distributed applications. This is a myth! It is the Internet, not the Web that allows for the deployment of low cost applications. All of the other so-called Web technologies have very little impact on the cost of developing, Table 2. Generic comparison of Web server-based and DICOM-based teleradiology systems Web server-based DICOM-based Cost Low Low Pull model operation Yes Yes Push model operation No Yes Internet interfaceable Yes Yes Point-to-point POTS Possible, but not Yes interfaceable generally implemented Standards based Yes (HTTP/HTML) Yes (DICOM) Compression Generally JPEG only JPEG and Wavelet deploying, or maintaining a distributed application, with few exceptions. Teleradiology is not an exception.

What drives the cost of any application is customer demand, the cost of engineering development, the cost of maintenance, and recurring operational expenses. This is the same for teleradiology. Only in the case of the simplest Web server-based teleradiology application does the Web result in a lower cost system. This system has price advantage in two areas: 1) the availability of low cost (even free) Web server software, and 2) the lack of programming required on the Web browser side. These advantages are engineering development advantages only. Maintenance costs, operational costs, and customer demand are not effected by these engineering advantages. It must be remembered how-ever, that the simple Web server-based system offers only limited features.

A more full-featured Web server based system requires significant programming effort on the Web browser side, if it is to compete in function and performance with DICOM-based teleradiology systems. In this case, the engineering costs are a nearly identical. Neither approach has any significant development cost advantage on the image viewing side. DICOM-based viewers generally have a per-formance advantage however.

The apparent cost advantage that Web server systems have, when compared with DICOM servers, is non-existent. Any medical imaging company must have DICOM technology suitable for the development of such a server. The ready availability of low-cost Web server software is meaningless to such a vendor. Their DICOM development costs have already been amortized over many other products, making the DICOM server technology no more expensive than Web server technology.

In the end, there is no difference in cost based on engineering development costs for a meaningful teleradiology application. Maintenance costs, recurring operating costs, and customer application demand are all the same.

Choosing between Web- and nonWeb-based systems

In order to decide whether or not to choose a Web-based system, it is important to understand the details of the application. In the end, the choice should come down to: which system addresses identified needs, as well as anticipated future needs, at the lowest cost. Table 2 provides a generic com-petitive analysis of Web server-based and DICOM-based teleradiology sys-tems. For the purposes of this compar-ison, the Web server-based and DICOM-based systems are assumed to have nearly identical functionality at the viewing end.

Conclusion

Although interesting and well hyped, the use of Web technology should not be the sole basis for decid-ing on the purchase of PACS and teleradiology systems. The technology employed in a medical imaging product should be one of many factors considered as part of the purchasing decision. In PACS and teleradiology applications, Web technology has no defacto advantage. As with any system purchase, no decision should be made without thoughtful consideration of the application itself and how well it meets the needs of the radiol-ogy department. AR

Dr. DeJarnette is the President of DeJarnette Research Systems, Inc., Towson, MD.