Applied Radiology

Many applications used in today's radiology departments require the use of digital networks. It is vital in teleradiology applications for both the in-house departments and remote sites. However, it is unrealistic to expect that the medical information systems department of your medical center is adequate to meet all of your institutions networking needs (the MIS staff has other significant problems, including implementing enterprise systems and dealing with disparate databases). It is, therefore, necessary for radiology department personnel to become knowledgeable enough to assist in coordinating networking efforts with the MIS department.

Radiology departments that have developed picture archiving and communication systems (PACS) have learned to work with both local area networks, or LANs (networks spanning distances of under 5,000 meters), and wide area networks (used for Internet applications). As is the case with any problem solving technology, new and improved network schemes are constantly evolving to streamline and improve workflow for transmitting patient data, including patient images. Many of these developments have been made quite recently and are capable of providing the bandwidth required for transmitting patient data (text and images) at a reasonable usage cost.

Radiology applications involving the proper selection of networks require an understanding of four concepts: bandwidth (analog and digital); the Internet and Web; signal-to-noise ratio (S/N); and the new networks of digital subscriber lines (DSL).

Bandwidth

The term bandwidth is a critical parameter in the choice of a network. For example, if a radiology department intends to transfer from a clinic to hospital 10 PA chest radiographs during an 8-hour period, then the requested bandwidth must be 16.66 kbits/sec (2,000 ¥ 2,000 ¥ 12 bits ¥ 10 images/8 hrs ¥ 3,600 sec/hr=16.66 kbits/sec). This assumes that the 10 films are transmitted in a uniform time interval during the 8-hour period. If the network's bandwidth is less than this, memory buffers must be used to queue the data so as to avoid losing the image data.

Data can be sent over a communication channel (i.e., transmitted on wires) by varying a physical property, such as voltage or current. This process is called signaling. The bandwidth of a given channel is a parameter used to define the amount of data which can be transmitted without distortion.1 The term continuous or analog signal means that a network is capable of transmitting a range of harmonics. A harmonic is represented by a mathematical function, f(t) = A cos (nwt), where "A" is the amplitude of the harmonic, "n" is the harmonic, and "w" is the frequency of the first harmonic. An ordinary telephone line, called a voice-grade line, has an analog bandwidth of approximately 3000 hertz (Hz), meaning that the highest frequency that can be transmitted is 3000 Hz, (3000 cycles per second).

In comparison to the analog type of signal, the term discrete or digital signal means a sequence of numbers defined for every integer "n". The sequence f(n) is called a digital signal and the index "n" is the discrete time. Thus, a discrete or digital signal is not represented by a continuous waveform but rather a sequence of values. In addition to quantifying time, a discrete signal quantifies the signal amplitude.

In measuring bandwidth, the rule of thumb is that more is better. That is, an analog channel of 1 MHz (1,000,000 Hz) is better than 1 kHz (1,000 Hz), and certainly much more costly. A digital channel of 1.544 Mbits/second (1,544,000) Mbps) is better than a 64 kbps (64,000 bps) digital channel.

The maximum data transmission rate of any channel was recognized in 1924 by Nyquist,2 who proved that if an arbitrary analog signal was transmitted through a channel of bandwidth B, the filtered signal could be completely reconstructed by making only 2 B (exact) samples per second. For example, a noiseless 3 kHz channel cannot transmit binary (two level) signals at a rate exceeding 6000 bps. If you try, distortions will occur in the reconstructed signals. This explains why modems used on telephone lines do not transmit past the 33.6 kbps rate and why the newest modems have difficulty reaching 56 kbps. In 1948, Claude Shannon extended Nyquist's work further to the case of a channel subject to random noise.3 Shannon found that the maximum data rate of a noisy channel whose bandwidth is B Hz, and whose signal-to-noise ratio is S/N, is given by:

maximum number of bits/sec = B log2 (1+S/N)

The signal-to-noise ratio measurement of a transmitting channel is similar to that of a radiograph. A good S/N is required in a radiograph to detect anatomic structures. For example, a channel of 3000 Hz bandwidth and a signal-to-noise ratio of 30 dB can never transmit much more than 90,000 bps. A transmitting channel with a good S/N is important to avoid errors in transmission. Public networks in which switched-links are employed (so-called dial-up lines) may not have as good a S/N ratio as that of a private network, one dedicated to your application.

Digital channels

Table 1 illustrates several digital channels and their bandwidths (data rates). Interface devices are used to transmit different types of signals. A modem is an interface device designed to carry digital signals (such as a text file) across a link which employs analog signaling. A Codec is an interface device required to carry analog information across a link using digital signaling. Another type of interface device, a DSU/CSU (data service unit/channel service unit) is required to change one form of digital signal to another.

The first invention to use this technology, the public switched telephone network (PSTN), was designed, engineered, constructed, and operated for the basic purpose of two people talking to one another for relatively short periods of time. Later, in the 1960s and 1970s, a 2400 bps modem was considered a high-speed device. In mid 1994, when the Web first surfaced on the Internet, the top modem speed was 9600 bps (9.6 kbps). Today modems signal at rates of 33.6 kbps, now pushing data rates of 56 kbps (at least in one direction: downstream).

An additional means of carrying a signal, the integrated services digital network (ISDN), has been a disappointment to the telephone companies. ISDN operates at 64 kbps, offers two channels, and has a primary rate interface signaling of 1.5 Mbps. ISDN is a switched service, meaning that an ISDN user can connect to other sites over the access line. This is important for teleradiology applications.

A T1 is a point-to-point, leased private line that can be provided at a fixed monthly rate (no user fee charges). It was developed in 1948 as a digital pulse-code modulation signaling system, signaling at data rates of 1.544 Mbps. Newer versions of this include the HDSL 2 (high-bit-rate digital subscriber line), and the ADSL (asymmetric digital subscriber line), which operates at a wide range of data rates (1.5 to 8 Mbps [downstream, uses one wire pair] and 16 to 640 Kbps [upstream]), and reaches to 18,000 feet (3.4 miles). A T3 line operates at 45 Mbps, as does an optical carrier level 1 (OC-1). Lastly, the broadband ISDN (B-ISDN) provides popular technologies, such as asynchronous transfer mode (ATM) and synchronous optical network/synchronous digital hierarchy (SONET/SDH). These high data rate transmission links are required for transmission of images between sites. High data rate links can transmit large sized data files with reasonable timeliness. Often, data compression methods are employed to reduce the size of the data files being transmitted in order to speed up the process. Many intended applications for these technologies have found their way onto the Web.

The Internet and the Web

The Internet started operations in 1969, funded by the Department of Defense's Advanced Research Project Agency; it was then known as the ARPANET. Its clinical growth was such that in the early 1980s ARPANET was divided into two parts, one for military networks (MILNET) and the other supported by the National Science Foundation (NSFNET).

The Internet is based upon a client-server architecture. The client-server model so often used in personal computers will maintain data on one or more shared file server machines, and the users of these machines are called clients. The Internet employs a file transfer protocol (ftp) standard, which is a client software package running on the client's computer and a ftp server software package operating on the server. Using this software, a home PC user client accesses a remote ftp server over the Internet and transfers an ftp file from the server to the home PC. Initially, it was difficult for the user to perform multiple applications, and worse yet, the user was required to learn lines of text commands to accomplish these applications.

Fortunately, a significant event occurred in the evolution of the Internet. Two approaches, known as hierarchical and hypertext, were tried to solve the problems of separate clients and text line commands. The hierarchical approach is based on organization and classification: a place to store everything and everything in its place. The first application to explore this approach was the Internet Gopher (developed at the University of Minnesota and released in late 1991). Gopher menus contained sites and resources from Internet locations that enabled users to "go for" them by interactive selection on computer screens.

In comparison, hypertext uses the relationship of the data as its guidance in moving about the Internet. Through hypertext, a user can move through lists of data across the Internet, going from one concept to another. The first hypertext application was a software package called HyperCard™ that was released by Apple™ Computers in 1987 as a part of the Macintosh™ system. Hypertext was brought to the Internet in the late 1980s by Berners-Lee and Robert Cailliau when they published their proposal for networked hypertext. It is in their proposal that the term "world wide web" was first introduced: "Hypertext is a way to link and access information of various kinds as a web of nodes in which the user can browse at will." The authors found a need to develop a language called hypertext markup language (HTML). As a result, Web software is now a mixture of Web servers (or Web sites) and Web browsers, which are software applications that operate on the user's client computer, communicating with the software operating on the Web server in order to transfer files.

As the power of PC computers grew, the requirements of a graphical Web browser were met, and the Web became the dominant presence on the Internet. Three fundamental events in the last 16 years have brought the Internet and Web to its current large existence: the personal computer (first developed by IBM in 1982); the creation of the Internet itself; and the development of a networked "Web". The next event projected to revolutionize the Web will be the digital subscriber line (DSL), intended to solve the problem of broadband residential access for advanced services.

Digital subscriber line (DSL)4

There are a number of possible communication technologies that can help with the problems of overloading in the public switched telephone network (PSTN) voice network with digital data and interactive broadband services, a common occurrence in telemedicine applications. If an entirely new system was put together, it would include satellite networks, wireless networks, and cable modems. However, the existing PSTN offers millions of copper-based analog local loops. On the analog local loop, frequencies above 4000 Hz can be used, which can be applied to the problem of overloading the PSTN with broadband digital services. It seems reasonable to apply copper-based solutions, if possible, to the issue of providing broadband services through available analog local loops. This is the purpose of DSLs.

The principal characteristics of the current digital subscriber lines are shown in table 2. This type of service uses a family of x-type DSLs. The asymmetric DSL (ADSL) is the service that is furthest along with respect to developed standards. It uses only one wire pair and reaches to 18,000 feet (3.4 miles). It is estimated that the cost of ADSL will be $200 per month plus the cost of an ATV (ADSL termination unit). ADSL employs ADSL modems, one at each end of the line. A splitter device comes between the local exchange and the customer's premises. This splitter allows existing analog voice telephone and other equipment to continue to function on the new lines; it also allows the long holding time data traffic to be rerouted around the PSTN voice switch onto an IP router or an ATM switch.

The xDSL is available here and now. It offers broadband services to and from radiology departments. Its cost is already decreasing due to the large market from the Web to residential homes. The xDSL family, especially ADSL, will significantly impact the radiology department's services (such as ultrasound). This family of products will reduce the cost of teleradiology and provide higher throughput data rates among radiology departments, hospitals, and clinics. AR

References

1. Stallings W: Data and Computer Communications, ed 4, pp 37-63. New York, MacMillan Publishing Co., 1994.

2. Tanenbaum AS: Computer Networks, ed 3, pp 81-82. Upper Saddle River, NJ, Prentice Hall PTR, 1996.

3. Derfler FJ, Jr.: Using Networks, pp 353-355. Indianapolis, Que Publishing, 1998.

4. Goralski W: ADSL and DSL Technologies. New York, McGraw Hill, 1998.

Dr. Dwyer is a Professor in the Department of Radiology at the University of Virginia Health Sciences Center in Charlottesville , VA. He is also a member of the editorial advisory board of this journal.