Is computed radiography (CR) simply a technological imperative — technology for the sake of progress? Powerful arguments can be made to support this stance. Implementing CR can be expensive and disruptive. It requires complicated equipment vulnerable to failure. Just storing and securing data becomes a major commitment of space, money, and effort. Why then, have so many succumbed to the temptation to acquire this technology? What are its advantages and how can it be implemented effectively?

The case in favor of CR, particularly in the emergency department (ED) environment, reflects needs within and well beyond emergency medicine. Targeting the ED for CR is a sound strategy to solve nagging problems with access to images for interpretation, the need to view images by multiple caregivers, and follow-up. However, implementing CR cost-effectively requires a sophisticated strategy to assure system usability and reliability and the ability to completely replace conventional radiography.

Because of its cost and complexity, CR makes the most sense for larger institutions, although it is also becoming practical for smaller ones as the price and size of CR units decline. The experience discussed here is from a large academic medical center with a level I trauma center. The ED has 20 beds in a 795-bed hospital. About 40,000 radiographic examinations are performed in the ED annually.

Designing A Strategy

Designing an appropriate strategy requires defining technical and operational objectives and then breaking the challenge down into its component parts. When beginning to plan for CR, a viable primary objective is to completely replace conventional analog radiology with CR. There are two key purposes to this objective. First, as images are to be made electronically accessible, if one does not completely replace analog imaging, then not all of the images are accessible, damaging the usability and credibility of this system. Second, failing to completely replace analog imaging would require maintaining parallel systems of imaging. This is costly, labor-intensive, and space-intensive.

However, as daunting as the deployment of CR is in the ED, planning for filmlessness is much more complex. Therefore, implementing CR should be seen as distinct from picture archiving and communications systems (PACS) or filmlessness, or at least as only one component of a PACS infrastructure. Nevertheless, using soft-copy reading and review is another valid objective of CR for the ED. Whether this soft-copy viewing is universal or partial must be based on the institution’s overall PACS strategy.

The need for some soft-copy viewing is based on the primary operational objective to eliminate the incidence of uninterpreted examinations. One embarassment of many busy EDs is that 5-15% of radiographic examinations may go uninterpreted, temporarily or permanently. The practice of clinical services removing films from the ED for urgent needs simply leads to their loss to radiologists and other health care workers. Soft-copy viewing can prevent this, but only if combined with adequate manual and computerized work-flow processes that assure that unread examinations can be identified and retrieved electronically.

Fault-Tolerant Architecture

Because EDs require nearly instant availability of radiological services 24 hours per day, to attain the objective of replacing analog radiography requires a highly fault-tolerant technical architecture. Because no single component of a computerized network of equipment can ever be considered fail-safe, redundancy of components is essential. However, in designing this redundancy, it is tempting to overengineer systems that back up and mirror every component of a system. Our approach has involved some compromises that have nonetheless been effective in achieving essentially continuous operation for more than 3 years without downtime.

The key components of a CR system involve the radiology information system (RIS), CR gateways, CR plate readers and cassettes, CR quality control (CQ) workstations, film printers, viewing workstations, network components, and the image archive. In our system, virtually all of these components are duplicated, except for the RIS, the network, and the archive. Even the network has an independent subnetwork that will permit continued operations in the event of some central network failures.

This duplication of components incurs additional expense, but returns its value in its high reliability. In determining the size of components, one should size each item to be able to accommodate full operation for a short period of time at full capacity, but not for extended operation. The complete, redundant system should have adequate capacity for growth. Additionally, the networking of the components should be designed to have fail-over switching should any one of the redundant components fail. Each of these major components also consists of subcomponents such as central processing units (CPUs), hard disks, and monitors. Spares of such crucial items are kept on-site.

A substantial challenge was simply finding the space to install these bulky new systems while analog capabilities were preserved during the transition to CR. This required some negotiation with ED leaders and promises to vacate space once the conventional darkroom could be eliminated. Coordination with institutional facilities planners led to a long-term, multiple-stage plan for work-space redesign. This included the eventual reclaiming of the old darkroom where some of the CR equipment was eventually relocated, away from clinical work space. Another space-related issue is the use of space protected from ambient light. If monitors must be placed in areas with constant overhead lighting, the use of higher brightness monitors is advisable.

Training Required

Upon installing the CR system, considerable training is required of technologists on all shifts to correctly process films and to avoid damage to the expensive cassettes. We required that only technologists who had been certified by our trainers were permitted to use the system. The character of both soft-copy and laser film images of CR differs substantially from that of plain films. Therefore, image quality parameters must be adjusted to be both technically satisfactory and subjectively acceptable to radiologists. This must be done for each major category of images, including chest, bone, and abdomen. In some cases, x-ray exposure factors must also be adjusted upward to adequately reduce mottle.

In our implementation, additional unanticipated problems initially arose regarding image annotation and orientation. For example, conventional cassette markers were not available to tag a particular exposure to a possibly damaged or dirty cassette. Such markers were designed by our informatics group in consultation with the technologists. These were fabricated by our machine shop and affixed to our cassettes. We also found that CR cassette exposure response varied from that of conventional cassettes. This led to a problem regarding lead-positioning markers. Technologists were accustomed to placing these markers just beyond the edge of the coned area where a faint x-ray shadow normally occurs. With CR cassettes, these markers often proved invisible either because of lower scatter exposure or because they were actually being placed outside the active receptor of the cassette. This problem was aggravated by the fact that it could not be quickly corrected with a magic marker on a film and if it was not addressed could lead to clinically serious confusion. These problems were corrected with process redesign and technologist training. Technologists were instructed by our trainers in person to affix the markers within the exposure area and in a standard location to assist the radiologists in orienting the images. Additionally, the quality control process allows capturing such errors, and these are corrected within the system when discovered.

A concern prior to installation was that there is a measurably increased time from exposure of a cassette to the completion of printing of a laser film compared to the printing of a film from a conventional processor. This time, in some cases, differed by as much as 1 minute per film. However, with two plate readers and stacking capabilities of four cassettes in each reader, this problem was minimized. This has not been a source of significant complaints.

Early evolution of our CR system has also involved marked work-flow changes in data entry. While at first, patient data were entered by hand, subsequently a bar-coded requisition was developed that permitted more accurate and rapid data entry. However, again, training and quality control monitoring are crucial to minimize the inadvertent use of medical record numbers for requisition numbers and vice versa. Eventually, the optimal solution involved the transfer of data directly from the RIS via HL-7 messaging. With compatible data gateways, an appropriate work-list of open requisitions can be presented to the technologist to select. This saves time and optimizes accuracy.

Each of the above subjects involves detailed work-flow issues with which technologists and radiologists are most familiar. We achieved effective improvements in this system only through the active involvement of key people throughout the design, implementation, and refinement process. Technologists, radiologists, computer scientists, and administrators all must be involved and consulted to identify the key problems and participate in solving them.

The Doe Issue

One problem that persists relates to the reconciliation of data when changes occur. In an active trauma center, many patients arrive unidentified and are given new medical record numbers and Doe names. When these people are properly identified, their new information must be properly tagged to the images. While the tools are available to perform this operation retroactively, it is time-consuming and awkward. Also, many such trauma patients have numerous radiographic examinations. It is not uncommon for some images from one series to be obtained under the requisition for another. In a conventional film environment, films can simply be shuffled and hung appropriately, while with soft-copy reading, it is more time-consuming to properly organize the review process.

This point emphasizes one of the frustrations of radiologists with soft-copy viewing. While films can be organized and hung by technologists or other assistants, with soft-copy viewing, this work is left to the radiologist, affecting efficiency. With attention to detail on the part of the technologists at the quality control step, problems with such image organization can be minimized. Because many patients in the emergency environment have numerous examinations, it is easy to inadvertantly, for example, place a foot image with a cervical spine study. If the technologist is careful in assuring that images are attached to the correct requisition, then when examinations are reviewed, the radiologist will not need to navigate among series of images to find the complete examination. While this process is emphasized in training for CR system use, periodic quality monitoring and feedback can improve performance.

Electronic image management can address another challenge to clinical physicians in an ED. It is often a frustrating task for an emergency physician just to keep track of when a patient’s radiographs have been completed. An automated notification system was designed within our own radiology informatics group that has become known as the Airline Monitor because of its resemblance to airport schedule monitors. This consists of a TV monitor in the central clinical area that displays each new patient examination at the top of the screen when it is completed. The oldest examination scrolls off the bottom of the screen. With a glance, one can determine if a radiographic examination has been performed.

Early Warnings

Incidents that occurred early in our experience were instructive. When we first designed and installed our CR system, before completing our transition, we attempted to economize by purchasing only one laser film printer, compromising our principle of using redundant components. Because of the heavy load on the printer, it experienced a mechanical failure. We found that in our zeal to convert quickly to CR from conventional processing, our technologists believed that our encouragement to use CR was instead a directive to use only CR at any cost. Therefore, although we had still maintained conventional processing at that time just for such a contingency, the technologists failed to use it, leading to a serious backup for several hours until the chief technologist was contacted. A second film printer was subsequently purchased. A second incident involved the collapse of a countertop overstressed by the weight of heavy monitors and computers. This downtime was brief and manageable.

Whether to convert completely to filmlessness depends on the resources of the institution to support widespread image access and the usability of the PACS that is selected for film interpretation. When we selected interpretation workstations about 4 years ago, only an older generation was available that was not well adapted to work flow or viewing for large volumes of cases. Therefore, we have continued to print films routinely for interpretation. However, although the PACS workstations are not optimal in quality, access to them has permitted us to interpret the approximately 10% of examinations for which the films are removed from the ED.

To minimize the time to generate film, we automatically print every image directly to film before any quality control QC process is performed. Therefore, we have not had a substantial decrease in film discard rates compared to a conventional analog process. If the process dictated that quality control measures be performed to determine if the image needs to be reshot (positioning) or modified (window/level) before filming, a reduction in film discards likely would result.

Access for interpretation thus also requires an ongoing method of storing the radiographic images using rapid access hard disk storage — RAID — and a long-term archive. This is the core of a PACS, which at our institution also encompasses all of our CT, MRI, portable radiographs, and those upright chest examinations that are performed with direct radiography.

Assuring the completeness of examination interpretation (without redundant reading) is managed by several mechanisms. Paper requisitions are used, and when films are removed, the requisitions are usually left behind. Therefore, a collection of paper requisitions without films becomes the initial PACS work-list. An Exception List is also generated each day, listing those examinations from the RIS that are not yet interpreted where the paper requisitions have also been lost. One of the most frustrating experiences of any radiologist is the rereading of an examination that has already been interpreted by someone else, an ever-present risk in a hybrid film-PACS environment. Such rereading is largely avoided in our institution through the use of a voice-recognition dictation system that, upon bar coding, instantly recognizes when a case has already been dictated.

Access Is Key

Although CR clearly is beneficial to the radiologist in providing the opportunity to access images at will, a further objective should also be the improved access to images by other clinical physicians. It is this that can lead to successfully achieving virtually complete filmlessness. The key to filmlessness is creating access to adequate quality and features of viewing wherever images are required to be viewed. If this can be accomplished even for a subset of images, then filmlessness can be attained for this subset. For example, if electronic images are accessible by orthopedic surgeons in their clinics, inpatient units, offices, and operating suites and to radiologists in their reading rooms, then routine filming of orthopedic images can be terminated without adversely affecting patient care. It is this accomplishment that can begin to provide the most dramatic economic payback for the investment in CR.

The issue of access involves not only access to radiographs obtained in the ED, but also to such studies as CT and ultrasound scans that might be obtained elsewhere in the institution. Electronic imaging permits ED physicians and consulting services to review all of the relevant data in a single location, improving care efficiency.

Also, multiple services frequently consult on patients. When a single set of films is the only way to view images, consulting services as well as radiologists are often deprived of access. While many PACS workstations are too complex for the casual user to master, Web-based image viewing programs are now available on ordinary PCs that can provide such access with little or no prior training. We have installed such systems in physically distant clinics, making the original images immediately available. However, despite the use of such systems, when films are available, one often finds that film is the usual preferred method of image viewing by those who only infrequently use electronic viewing.

A recent study in our ED shows that we have accomplished the difficult transition from analog film to CR without measurably affecting the length of time patients stay in the ED. By overlapping conventional and computed radiography, only one major disruption lasting several hours has occurred in more than 3 years. Access to images by radiologists and clinicians has been markedly improved and the rate of unread examinations has been eliminated. While the implementation of CR in the ED should not be taken lightly, it can provide great value both in the quality of care and as a basis for the cost savings and improved productivity to be achieved through the eventual complete conversion to electronic image management.

Lincoln L. Berland, MD, is professor and vice-chairman for administration and planning, Department of Radiology

Kevin L. Junck, PhD, is associate professor and chief of radiology informatics, Department of Radiology, University of Alabama at Birmingham.