The PACS revolution has been made possible by the dramatic proliferation of computers and associated technologies. The forces that have commercialized and provided the Internet to business, as well as the general public, also have contributed to the development of PACS for radiology. We live in an electronic age where "information" is accessible digitally through the use of computers and communications networks, and radiological images are just another type of information that can be manipulated, distributed, and viewed by computer. Direct digital images such as MR or CT can easily be transferred from computer to computer, within a hospital or across the world to a colleague for consultation.

Computers --Computers are the most obvious elements of a filmless system. A computer is required to collect and store the electronic signals generated by a MR or CT scanner. The computer then performs calculations in its central processor unit (CPU), converting these signals into useful images. The human interface with a computer takes the form of a keyboard, mouse, and video monitor(s). The speed of computers has doubled approximately every 18 months since their inception and is characterized by the CPU clock frequency (speed). Computer speed is measured in MHz (i.e., millions of cycles per second), and processors found in current PCs operate at about 450 MHz. This number is a measure of how many computations, such as adding two numbers, can be performed in one second.

The CPU in a computer implements the instructions of a "computer program," performing calculations or displaying images on a monitor. All communications with a computer are ultimately in the form of numbers. Strings of numbers make up the computer program, the image data, and the entries from the mouse and keyboard. These numbers are stored in the computer's random access memory (RAM), where the CPU has access to and can work on them. RAM is characterized by how much information it can hold and is reported in units of millions of bytes (MB). Most personal computers come equipped with at least 32 MB of RAM. This number affects the computer calculation speed, as well as the ability to rapidly manipulate images on radiology viewing stations.

Information in RAM is erased when the power is turned off to a computer. Digital data (i.e., programs and images) which need to be retained for any length of time are stored in a more permanent fashion on a magnetic hard disk in the computer. These devices hold
billions of bytes of data, with disk sizes of 8 gigabytes (GB) being common. For archive purposes, images are ultimately stored on cheaper media, such as magnetic tape or optical disks. The amount of image data generated by a typical radiology department annually is of the order of thousands of GB (see below).

Image displays --Size, resolution, refresh rate, and luminosity characterize the video display (monitor) used to view images. The size of most radiology monitors is 19 to 21", measured on the diagonal. The resolution of the system is how many pixels horizontal and vertical it is capable of displaying. Most computer/monitor systems are capable of displaying at least 1280 * 1024 pixels, and could display four CT (512 * 512) images at one time with no loss of information. Refresh rates should be greater than 70 Hz, otherwise flicker results which is fatiguing on the eyes when viewing images. There are some very expensive high luminosity monitors made especially for radiology and these are expected to be helpful in high ambient light conditions.

The term workstation is often used when talking about "higher-end"
computers. These run a Unix operating system, are more expensive, and incorporate higher performance components, such as very fast disk drives and fast video drivers. The distinction between personal computers and workstations, however, has become fuzzy. For example, the performance of a personal computer today surpasses what would have been called a workstation just a few years ago. However the technical requirements of radiological image display systems are still being actively debated, as the cost/performance ratio continues to drop rapidly.

Networks --Computers can be made to communicate with each other using an electronic interface and special wiring. The network interface card (NIC) provides this capability. Of the variety of communication standards, Ethernet is the most common. A typical Ethernet adapter communicates at ~10 million bits/second (10 Base-T), using twisted pair wiring with connectors similar to telephone wiring. Higher transfer rates of ~100 million bits/s (100 Base-T) are now possible, but these require a better grade of wiring and faster networking components. The speed of the disk drives and the characteristics of the network connecting any two machines will determine the transfer time for an image.

This discussion of network consideration so far has described the workplace environment. Teleradiology between the workplace and the home is also possible and introduces other considerations. Transportation of digital information in this situation is often via a telephone line--an inherently slow technology. New digital transport technology is now becoming available in the form of integrated services digital networks (ISDN), satellite service, cable modems, and digital subscriber lines (DSL) (table 1). ISDN and DSL are services provided by many local telephone companies over phone lines. These services are relatively expensive and frequently have an additional fee based on connection time. However, such services provide rapid data communication to computers at sites that generally are only serviced by telephone lines. T1 service (also usually delivered from the phone company) is an expensive utility relying on dedicated wiring to provide fast communication between computers at distant geographic locations, or between a computer and the Internet.

Recently, satellite and cable TV companies have gotten into the act, providing inexpensive connections between home computers and the Internet. Most often there are packages with a fixed monthly fee independent of connection time. The cost and availability of these technologies is becoming more favorable as the Internet expands its hold on our lives. If security issues are properly addressed, the Internet seemingly offers the most promise for transmitting digital information. At our institution we have used a T1 line connection to the Internet to implement an inexpensive teleradiology system, with cable modem access for physicians at home (information from: http://alpha. nmrlab.hscsyr.edu/nmr_lab/asnr98).

Digital image information --The amount of space used to store a digital radiological image depends on its matrix size and the number of shades of gray (or colors) present in the image. Conventional radiography, such as a chest x-ray image, requires about 10 MB of storage space per image, whereas a nuclear medicine image requires about 1/1000 storage space (10 KB) (table 2). In more familiar terms, a chest x-ray would require 7 floppy disks to store it, or one could fit 65 such images on a common CD disk.

Table 3 shows the workload of a (hypothetical) radiology department which performs 150,000 radiological examinations per year (~500 patient examinations per day). Included is a typical breakdown of the number of examinations performed in each of the nine major areas of a radiology department, together with a representative value of the information content of a typical patient examination. Examinations associated with large information contents include mammography, special procedures, and fluoroscopy. About 60% of the total 3,200 GB of information generated in this department is obtained using screen-film radiography. It is of interest to note that mammography, which accounts for only 2% of the patient examinations in the department, generates about 20% of the total information content.

Digital modalities --Computed tomog-raphy (CT), magnetic resonance (MR), and nuclear medicine (NM) are inherently digital imaging modalities. Ultrasound also can be made a digital modality by using newer equipment incorporating digital electronics, or in older equipment by the addition of a frame grabber, which digitizes the video signal being fed to the monitor display. Imaging companies are currently modifying these digital image acquisition modalities to ensure that the output data conform to the DICOM (digital imaging and communication in medicine) standard, making images compatible with and readable on nearly all manufacturers' systems. The adoption of the DICOM standard will help to facilitate the transfer of images from acquisition devices to workstations, printers, and archives.

Special procedures, fluoroscopy, and cine are increasingly providing a digital output by digitization of the TV signal. Today, any newly installed TV/image intensifier system should be digital, thereby maximizing image quality and improving operational efficiency. Digital mammography also is currently being evaluated by several companies and will be commercially available in the near future. However, the large mammography image data size and the associated problems of displaying this large amount of pixel information (as well as the cost of the equipment) are major hurdles that still need to be overcome. For the immediate future, therefore, it seems unlikely that digital mammography will replace screen-film mammography in most clinical settings.

Despite the advent of digital imaging technology, the overwhelming majority of radiographic examinations today still are screen-film radiographs (table 3). Photostimulable phosphor systems, also known as computed radiography (CR), were first introduced in the early 1980s as a replacement for screen-film combinations. These systems, however, still suffer major limitations, including operational inefficiencies relating to the need to manually handle the cassettes, and a poor imaging performance due to excessive light scatter in the photostimulable phosphor.

Alternates to screen-film --A viable and cost-effective alternative to screen-film is an essential prerequisite for most radiology departments before they will be able to convert to an all-digital mode of operation. Charge couple devices (CCDs) capture light (from screens) and convert this into an electrical (digital) signal. The large phosphors are coupled by optical lenses to very small CCDs, resulting in a very low efficiency for capturing the light produced by the x-rays absorbed in the phosphor. Photoconductors such as selenium may be used to convert x-ray photons directly into an electronic charge. The stored electronic charge may be read out directly, using a variety of read-out methods, and has demonstrated an excellent intrinsic spatial resolution performance. Dedicated chest systems are now commercially available but are relatively expensive. Flat panel detectors based on selenium are becoming commercially available, but the price of these systems will likely determine their ability to replace screen-film.

Indirect capture is the most promising digital technology likely to replace screen-film in the near future. Indirect capture refers to the use of a conventional phosphor (e.g., cesium iodide) to capture x-rays and convert a fraction of the energy to light. An electronic array is placed adjacent to the phosphor to capture the light emitted by the phosphor and converts this into a (digital) electronic signal. The x-ray image will be available for review within seconds of the exposure, eliminating the need for chemical processing. It seems likely that flat panel devices will be used for both radiography and fluoroscopy, eliminating the need for bulky image intensifier systems. Introduction of cost-effective indirect capture devices as a replacement for screen-film will clearly help to facilitate the move to PACS.

Hardware --The hardware required to convert a conventional radiology department to PACS is generally quite expensive. For example, the list price of a two-monitor "high resolution" display system is of the order of $120,000. A department is likely to need one workstation per faculty within the radiology department for primary diagnostic work. In addition, a large number of cheaper (lower quality) display stations will be needed to permit images to be displayed in clinical areas (e.g., intensive care units and the operating room) and to referring physicians.

Images acquired by the radiology department will need to be stored in several "short term" archives for periods of about 3 months before being migrated to a "long term" archive. Local storage will likely have about five 100 GB archives, each costing about ~$50,000. A "long term archive," such as an optical juke-box, will be expected to have a capacity of the order of 7,000 GB and will cost $180,000 for the actual hardware, with another $200,000 for the database managing system. Also required is a network to link the acquisition devices, such as CT scanners, with the archives and
image display stations. In addition, an efficient (and intelligent) software system will be necessary to ensure that the acquisition, transfer, display, and storage/retrieval of image information can take place in a timely and efficient manner.

Table 4 lists the major components of a PACS system, with the approximate (list) prices from a major radiology vendor. Current prices are clearly related to performance, and effective integration of the disparate items of PACS-related hardware is not trivial. PACS hardware will need to be maintained and upgraded periodically, which will result in significant ongoing operational costs. Additional factors to be considered include the training costs for the users of PACS systems, and ensuring the availability of service personnel to rapidly respond to problems associated with the successful retrieval and display of images 24 hours per day.

Software and integration --It is hard to overemphasize the role of software in an efficient filmless department. In
addition to being reliable and expandable, an effective PACS system must be user-friendly to the radiologist, allowing him/her to work with at least the same efficiency as with film. An important and often unappreciated factor is the efficient integration between the PACS system and the radiology information system (RIS).

A fully integrated RIS will communicate with the PACS system initiating the retrieval of prior images from the archive so that the radiologist will have the necessary information at the workstation to read the study. If RIS information is transferred directly to the scanner, there is less chance for error in complying with physician's orders, as data entry is greatly simplified. Images also can be sent to the hospital information system (HIS) and be made available as part of an "electronic patient record". A HIS/RIS interface is required to integrate the patient data base with viewing software and is expected to cost of the order of $150,000. It is likely, however, that the costs associated with this integration will be subsidized in part by the institution's information technology (IT) group.

Cost savings --Film is one of the largest savings within a filmless department, as it is a major expense in radiology. A department performing the workload depicted in table 3 is expected to spend on the order of $1 million (list price) for film. Current discounts from major film vendors, however, can easily be 60% to 70%, which makes the financial justification for PACS difficult. On the other hand, elimination of chemical processing will result in substantial benefits in terms of reduced waiting times for images (e.g., in GI studies), reduced costs, and reduced chemical waste that would otherwise have to be processed by the radiology department. PACS also will result in some manpower savings. Some jobs, such as the file room clerk, will likely disappear, whereas reductions would occur in other types of jobs, such as the technologists operating the imaging systems, due to increases in operational efficiency.

It is, however, important to note that there will be a need to hire relatively higher salaried individuals (systems administrators, etc.) to operate the PACS system. In addition, there will be significant operational costs to maintain the PACS system (service engineers, quality control technologists, systems trainers etc.). On current assessments, the installation of a PACS system will
be very costly. Today it seems clear that the major benefits of PACS will be the improved efficiency of the delivery of radiology services to the referring physicians, and not major dollar savings by the radiology department.

Archive/retrieval/viewing --Today, many facilities already archive to magnetic tape or optical disk in addition to printing the films, in case original quality images are needed at some future date. Digital archival will eliminate the need for bulky film file rooms and off-site storage. Lost films also are eliminated by the use of digital archives. These digital archival systems enable the rapid and convenient retrieval of old studies by multiple users. In addition, stored digital images on optical disk do not degrade with time and can be generated for legal considerations as well as for medical applications.

Digital image availability will result in significant improvements in radiologists' ability to efficiently read patient studies. Examples of this include the ability to perform cine loop viewing for angiography and cine studies. Additional benefits to the radiologist include the ability to adjust the display of any image (window/level), as well as the efficient retrieval of prior images for comparative side-by-side viewing. Of particular importance will be the ability to imbed images within the report for the referring physician, as well as to provide the reports associated with images viewed at remote display stations by users of the radiology imaging services.

Image transmission --PACS permits the convenient and prompt dissemination of radiographic images. Images may be forwarded directly to clinicians (e.g., bedside chest radiographs for ICU physicians). Digital images also may be included in an electronic patient record for convenient review by any physician or medical practitioner. These images may be sent to central locations for review by colleagues or a sub-specialist. A key issue here is the speedy provision of verified reports, which is the most important end product of a radiological examination. The provision of a final radiographic report is likely to be revolutionized by the imminent availability of reliable voice recognition systems.

PACS will naturally result in the provision of teleradiology for on-call or other home access purposes. This concept could be extended to permit "taking a call" from another time zone to allow work during normal hours. Other benefits include the provision of flexibility in servicing multiple (satellite) locations, as well as the availability of specialist over-reading capabilities for the general radiologist. It is likely that the greatest impact will be on the users of radiology services, i.e., the referring physicians, who will get access to images and reports in a more timely manner.

Image processing --The availability of digital images permits enhancements to image visualization as well as image quantification. Image processing options currently available include tonescale enhancement (via window/level manipulation) and other forms of image filtering to enhance specific features in radiographs. Reformatting images in other planes is possible and may provide benefits in certain applications. Other examples of imaging processing options include 3D rendering for surgeons and image fusion, where functional images obtained from nuclear medicine might be combined with anatomical images obtained from MR or CT.

It is important to note, however, that many of these image processing techniques are currently quite time consuming and could slow down the interpretation of radiological images. Image processing could eventually improve overall diagnostic performance, as accuracy for a given diagnostic test and the speed of these calculations is continually increasing. For the present, however, the costs of (any) increased reading time would need to be evaluated in the light of any corresponding improvement in diagnostic accuracy.

There is no reason why images cannot be shared and archived as easily within a radiology department as a document or illustration is handled in a corporate setting. This statement can now safely be made due to the staggering advances in networking and personal computing technologies. It is common to take a 20 MB computer-generated slide show to a business meeting and project it using a small video projector. A CT examination is about the same file size. Twelve years ago, our research group had to transfer images from a MR scanner to another computer using a removable 16 MB hard disk. The system was upgraded to enable electronic transfer of data to a computer workstation via an Ethernet adapter costing $10,000. Today, the same Ethernet hardware comes on a single card which plugs into a personal computer, works 10 times faster, and costs less than $100! Similar stories apply to disk storage and computer memory, emphasizing the fact that there has been exponential growth in the capabilities of computer and network technologies while prices have plummeted.

PACS, in general, will require an investment in the radiology department for the benefit of the medical institution served by radiology. Several hospitals have successfully become filmless, including the Hammersmith Hospital in London, OMSZ Hospital in Vienna, the VA Hospital in Baltimore, and the Medical University of South Carolina (MUSC) in Charleston. The benefits of PACS are significant. PACS eliminates lost films and improves the overall efficiency of a radiology department. The most direct beneficiaries of PACS will most likely to be referring clinicians rather than radiologists per se, with the radiographic images being ultimately available through a totally electronic patient record. There is little doubt that all radiology departments will eventually embrace PACS, although there is still much ongoing debate as to the "how" and "when" of this move to filmless radiology. AR