Last year it appeared that digital radiography could become the dominant mode of radiography, if not by the year 2000 then shortly thereafter. And while attempts are under way to purge film, technological complexities, cost, and regulatory wrinkles are slowing installation, particularly for the 70% of US hospitals with fewer than 200 beds.

Decisions in Axis Imaging News wanted to know what has been learned recently about installing computed radiography (CR) and direct capture radiography (DR) (for a discussion of the differences, see sidebar). We asked what clinical and financial benefits had been obtained, what problems had been encountered, and how users see the technology’s role today and in the near future. We spoke to Theron Ovitt, MD, professor and chairman of the Department of Radiology at the University of Arizona, Tucson; Master Sgt. Rik R. Guinther, director of the Department of Radiology, Brooke Army Medical Center (BAMC), Fort Sam Houston, Tex, which since 1995 has used film only for mammography (originally, BAMC installed a CR system, but for the past 6 months, it has been evaluating the amorphous selenium technology); Jonathan Tucker, also with the radiology department at BAMC; Etta D. Pisano, MD, associate professor of radiology and director of the Breast Imaging Section at the University of North Carolina, Chapel Hill, and a pioneer in the field of digital mammography; and H.K. Huang, DSc, vice chairman of the Laboratory for Radiological Informatics (LRI) at the University of California, San Francisco (UCSF). These experts offered numerous observations and tips for anyone considering a conversion to electronic imaging.


The obvious first question is whether radiologists and clinicians are satisfied with the quality of the images. Ovitt pronounced himself very satisfied with image quality on his CR system. “The limiting resolution is about 2.5 line pairs/mm, whereas film is about 5 line pairs/mm. That difference is hard to distinguish. At the spatial frequencies where we normally operate, 0-2 line pairs/mm, the two modalities are equivalent.

“Overall, the image quality is better with CR than with film because you have much more control over latitude and contrast on a CRT (cathode-ray tube) viewing station,” he continues. “You can present an image in a way that highlights what you want to see. For example, in the chest, you can adjust the display to highlight the mediastinum or the lung.”

Tucker is doing an image quality comparison of CR and amorphous selenium technology using a new phantom, in which disks with various sizes of embedded objects are positioned randomly. “We arrange the objects, image the phantom, and have a radiologist hunt for all of the objects. The radiologist is told only how many objects there are. He or she then lists the objects in order of obviousness. What we found in our preliminary testing is that radiologists can find more of the objects on DR images than on CR images,” Tucker says.

Tucker made two significant points about image quality and its assessment. “We have had the DR system for only a few months, and it has not been operational all that time. Therefore, we probably do not have it optimized yet. Also, both our DR and our CR systems capture images at 12-bit depth, whereas our picture archiving and communications system (PACS) can handle only 10-bit images. Thus, the images must be compressed before they are viewed by the radiologist. When we look at the DR images at the full 12-bit depth on the vendor’s workstation, we can see much more. Presumably, the same would be true of CR if we had a way of viewing them as 12-bit images.” Software that will upgrade the BAMC PACS to 12-bit images has just been released, “but installation probably will take more than a day, so we have to choose that time very carefully,” Tucker remarks.

Digital mammography still has not been approved by the Food and Drug Administration, but Pisano, who is conducting some of the clinical trials, says she is very enthusiastic about the equipment she is using. “I am optimistic about it being useful in finding more cancers, but until we have the final trial data, I cannot be dogmatic,” she says.


Many filmless pundits have stated that it is difficult to calculate the true costs and savings of electronic imaging. An added consideration is that the costs are incurred immediately, whereas the savings from such outcomes as faster turnaround of reports and more efficient use of the radiologist’s time may not be manifested for years.

Training is a significant cost issue: training not only of the radiologists, technologists, and clinicians, but also of the support staff, as everyone’s job will change when the new system is installed. For example, file clerks become database managers.

Nor is mastery of the new technology the only training issue. Training must continue in film-screen procedures for safety net purposes, according to one veteran. “If your electronic system fails, you will have to go back to film,” Guinther warns. “You need to be sure you keep a supply of film that is not expired, that the technique charts are appropriate for the equipment available, and that the technologists know how to use that equipment.”

Installation, service, and maintenance costs are another significant issue and differ for CR vs DR systems. Converting to CR imaging requires only the addition of a plate reader, which can serve several rooms, and the plates for the film-screen cassettes. The situation with the direct capture technologies is more complex.

“Yes, DR can be retrofitted into present equipment, but it is not as easy,” Tucker notes. “The vendor of our DR system took one of our examination tables back to their factory, removed the bucky tray, and installed the detector array. This process entailed raising the table to accommodate the tray and alterations in several mechanical features. For example, they had to counterweight the table so it can be set on end to obtain chest films. Also, the detector array does not exactly duplicate the bucky tray, and there are some studies, such as cross-table examinations, for which we must still use CR. Also, patient positioning over the array is sometimes difficult.”

Tucker also points out that at present, if some of the detectors in a direct array fail, the array must be fixed or replaced. “With CR, you simply replace the plate.” He suspects that repairs will become easier with time, but at present, maintenance can be a significant issue.


User support costs must also be considered. Radiologists tend to be enchanted with the technology of computer-based imaging systems, but the clinicians generally are interested only in looking at the pictures. If the new digital imaging system proves intrusive, they may view it as radiology’s bothersome new toy.

“The radiology department operates the PACS and provides the service, but CR, DR, and PACS are a hospital system, not a radiology system,” Guinther stresses. “The clinicians may say, ‘We are happy with film. Why do we need this monitor to give us the same images but that requires us to log on and log off?’ You have to show them how the new imaging can benefit them.

“If I am going to sell you something, I must provide support,” Guinther continues. “If the clinicians call because they have forgotten the password or cannot gain access for some other reason, and nobody is there to solve their problem, they get very frustrated with the whole system and do not want to use it.”

The savings from the dramatic reduction of film usage can be sabotaged if there is large-scale printing of images.

“We have several residency programs at BAMC, and people are always trying to print images for their teaching files,” Guinther says ruefully. “To prevent indiscriminate use, only a few people know the passwords that allow images to be printed.

“Essentially, the only time we print a digital image is when it is required for special circumstances such as in the operating room or one of our patients is being sent to a hospital to which we are not connected digitally.”

Some early writers on filmless departments suggested that a hospital might want to digitize all of its archival files. This is a costly and time-consuming approach, especially when one remembers that most of the images in the archives will never be viewed again. At BAMC, films are digitized only if a patient needs to take one to another hospital. “We digitize that image so we will still have it if the film is not returned.”

The cost considerations of CR and the direct capture technologies differ according to the size of the workload.

“I do not know if there are any large cost savings for CR technology for medium-size or small operations,” Ovitt says. “The main advantage of CR in this situation is the bridge it provides into a PACS. For very high-volume operations, on the other hand, direct digital capture is cost-effective.” Ovitt believes that CR will replace film-screen imaging at medium to large hospitals in the near future.

“PACS is now affordable,” he points out. “I think it is going to take off right after we finish worrying about Y2K!”


One of the arguments in favor of digital imaging methods is their facilitation of distance diagnosis (teleradiology). Most of the PACS now being sold incorporate Web-server technology that allows images to be broadcast over wide networks. Does this feature affect cost considerations for smaller hospitals?

“Probably not,” Ovitt says. “Teleradiology is cost-efficient for urban hospitals where the physicians serve multiple other sites or there is a large urgent care/emergency department that needs to have immediate consultations with referring physicians. But even with the smaller hospitals we serve, we find it advantageous to have a radiologist on site. We are available for backup, such as on nights and weekends, and for specialty consultations, but few small hospitals are doing complicated head and spine studies or MRI that they need to send out for interpretation.”

In other words, Ovitt does not see digital imaging and teleradiology leading to the sweatshop operations predicted by some observers in which radiologists sit before monitors all day, interpreting images sent electronically from hospitals to far-flung geographic areas.


Mammography is simultaneously a particularly desirable and a particularly difficult field for digital imaging: desirable because the ready availability of long-distance transmission of the images for review would improve delivery of screening, particularly to rural areas, and difficult because lesions as small as 1 mm are significant and thus must be identified. It seems that digital mammography has been on the horizon for several years.

“The FDA changed its approval criteria in February,” Pisano reports. “Originally, it asked for agreement studies demonstrating that radiologists agreed on the interpretation from film-screen and digital images, and a trial of this type by the International Digital Mammography Development Group is nearly completed. Now, however, the FDA is asking for correlation with truth. That is, it wants us to show that if a patient has a breast biopsy, digital images predict the outcome of that biopsy. I understand their caution: it would be tragic if after 10 years of digital mammography, we found that the mortality rate from breast cancer was rising. But it would have been nice to have the approval criteria established before we started our clinical trials.” The International Digital Mammography Development Group is completing the analysis of the data from such a trial. (The FDA still has not issued its criteria, and individual manufacturers are consulting with the agency about protocols.)

Another promise of digital imaging is the ability to harness computers to speed the reading of films by making a preliminary evaluation of an image for abnormalities (computer-aided diagnosis).

“When you view a CR or DR image at different parameter settings, that is a form of computer-aided diagnosis,” Ovitt notes. “But automated computer-aided diagnosis (CAD) is still a work in progress. Your neural network for image interpretation is still inside the radiologist’s head.”

Pisano is sanguine about computer aided diagnostic digital mammography, which she believes will be approved by the FDA within a year or two of the clearance of digital mammography.


The telemammography project at UCSF is supported by the California Breast Cancer Program, the National Library of Medicine High-Performance Computing and Communications Program, and the US Army. There are two digital mammography image-acquisition machines, one at the Ambulatory Care Center at the Parnassus campus of the university and one in the Breast Imaging Section at Mt Zion Hospital about 2 miles away. They are linked to each other and to LRI by an asynchronous transfer mode (ATM) SONET ring that moves the images at 632 MB/second. Images are interpreted on 2500-line workstations built by the LRI. A direct digital mammogram can be acquired in about 6 seconds for each of the four images, and the images are available for viewing 2 minutes later.

An invaluable feature of these workstations is their two cursors of different colors. “I can move my red cursor to a lesion, and the red cursor also moves on your monitor, and we can discuss the findings,” Huang reports. “Or you can say, ‘What about this?’ and move your blue cursor, which will also move on my monitor. So we can discuss these images easily in real time.”

At the moment, this telemammography project serves only two hospitals linked to LRI, but next year, there are plans to extend it to Stanford University.

“The distance does not matter if you have compatible equipment,” Huang points out. “The problem is associated with the carrier you are going to use. If you transmit a mammographic study using the Internet and regular telephone lines, it might take you 3 hours. The next-generation Internet now being developed will have almost the same speed as our ATM SONET ring.”

Security is an issue whenever medical data travel over public lines, and the LRI has taken two steps to protect its images.

“We worked with the mammography hardware manufacturer so that before the data go out onto our network, they are encrypted automatically,” Huang says. “But there is another consideration, and that is, what if someone tampers with the image while it is in transit, and you do not know that? So we wrote a software program that produces a digital signature. Every digital mammogram has its own specific signature. If somebody changes even one pixel, we will know the image has been changed and where, although we will not be able to tell how it was altered.” This program has been written to be usable with the next-generation Internet.

Clearly, a hospital considering a conversion to electronic imaging must understand the various technologic options and their advantages and disadvantages. Numerous parties need to be involved in the planning and implementation, and they all should be prepared to make realistic compromises. Once the difficulties are overcome and the system is running, the hospital will find that no one wants to go back to the good old days of film.


Judith Gunn Bronson, MS, is a contributing writer for Decisions in Axis Imaging News.