Digital Radiography

article 5
article 1
article 2
article 3
article 4


DR's power requires networking to match

By: Douglas Page

Pure digital radiography can make imaging suites more efficient but is not yet available for portable or bedside applications.

With their rapid digital processing and ability to route images automatically, digital radiography (DR) systems have irresistible appeal. Image cycle time is reduced to the range of 35 seconds. Image review is theoretically possible within 10 seconds of exposure. DR systems usually require fewer retakes, due to their extended exposure latitude, and patient radiation dose is potentially reduced, thanks to fewer repeat exposures.

Proponents of DR also claim that it simplifies workflow and enhances productivity, enabling staff to conduct more exams in the same time period. But DR comes with a dowry of concerns about PACS workflow, patient throughput, and network connectivity that require careful consideration.

"The most important issue is to ensure that the patient exam information and demographics can be combined for both DR and CR image sets," said Larry Filipow, Ph.D., an assistant professor of radiology and diagnostic imaging at the University of Alberta.

One radiographic examination commonly produces several "in-bucky" DR views, and then a few "out-of-bucky" or "crosstable" views using CR plates.

"Now you have a situation where, for one examination, you have both DR and CR images produced, most likely by equipment from different vendors," Filipow said. "These images have to be filed together as one 'visit' or event."

Because most PACS are not yet capable of amalgamating these events as one radiographic exam, the result is a confusing exam composed of two files created from one patient visit-one with DR images, one with CR images.

While the future of all diagnostic images is unquestionably digital, new developments in imaging modalities have made it difficult for PACS to accommodate the bloated requirements for storage and image retrieval.

One principal characteristic of digital imaging modalities is the drastic increase not only in the number of images they produce but in the amount of data per image.

"A synergism has been created in which the ability to retrieve and display very large image studies has led to the development of imaging equipment that produces still larger data sets at faster rates," said Robert Gould, Sc.D., a professor of radiology and bioengineering at the University of California, San Francisco.

A single digital mammogram, for example, can exceed 30 MB. A CT chest study can create 140 images in less than 20 seconds, amounting to 73 MB of data. Biplane rotational angiography machines using image intensifiers are able to create between eight and 32 images per tube rotation, easily generating 500 MB per study.

It's no different with DR. A single uncompressed DR chest image, at ~2.5K x 3K pixels, with a bit depth of 12, will inhabit 15 MB of storage space. Add a second view, and one routine uncompressed DR chest examination requires about 30 MB of storage space.

Practically speaking, network bandwidth is only 30% of stated nominal speeds, so the time triples to 75 seconds or more. But with automatic lossless 2:1 image compression, the actual transmission time drops to somewhere around 30 to 40 seconds for a 10Base-T (10-Mbps) Ethernet network, he said.

"This is too long in a busy clinical situation, so the PACS will need to prefetch previous exams, and autoroute the current exams to the radiologist's workstation to achieve clinical acceptance," Filipow said.

A second option is to ensure additional bandwidth, with a 100 Base-T (100-Mbps) network, that enables effective transmit times of about three to four seconds, usually acceptable for ad-hoc image/file queries.

"For LAN networks, 100Base-T is minimum," said Henri "Rik" Primo, director of medical solutions in the IS/PACS CN department at Siemens Medical Systems. "Gigabit is better for a busy department, but if a DR was installed in a remote imaging center, with low throughput or low exam volume, routing over a wide area network with a standard T1 line to the central facility could work."

A PACS that was designed for ultrasound, CT, and MRI does not allow a department to add general procedures and/or CR and DR without substantial resizing and enhancements, including CPU capability, RAM size, RAID, and archive capacity. And it may become difficult or impractical to work with optical disks for long-term archive (LTA), or to expand the existing optical jukebox, Primo said.

"This may be the ideal time to upgrade the LTA and migrate to tape archiving," he said.


The courtship of DR by radiology departments has provoked a number of debates, among them a dispute over the optimal viewing configuration for DR chest studies-the ultimate challenge for a PACS, according to Filipow.

To view the full image in "native resolution" (one display pixel for each acquisition pixel) requires a 3K x 2.5K portrait monitor, and to view both posteroanterior (PA) and lateral views requires two of them. If the radiologist wants to view the previous PA and lat exam, add two more monitors-each of which is prohibitively expensive to buy and maintain, and lasts three years at most. And viewing CT, ultrasound, or MRI images on one of these monitors is next to impossible, Filipow said.

Images in native resolution are too small, and need to be enlarged (4 or 16 display pixels for each acquisition pixel), he said. The alternative is to view the chest images on 1.5K monitors at reduced resolution (1 display pixel per ~3 to 4 acquisition pixels), and if necessary, to magnify regions of concern on the image. A less expensive alternative, these monitors can be used to view CT, ultrasound, or MRI images in native resolution so they are brighter and last longer.


DR has many of the same issues as other modalities, Primo said. It is essential that facilities take advantage of DICOM worklist features and RIS feedback. Other experts say patient throughput is an important issue, since patient traffic may increase at single DR modality rooms.

When thorax exams are performed in the film-based model, patients are distributed over many rooms, as almost all rooms have a vertical bucky, Primo said. The same workflow model can be used with centralized CR.

"Although the patient throughput of DR is great, there may be a traffic jam to access DR through the typical two dressing rooms that come with one rad room," he said.

Possible solutions include installing more dressing rooms, routing patients to a second DR room, or continuing to use CR in all rooms.

Quality assurance programs also need to address the presence of DR.

"DR may be very forgiving for under- and overexposure parameters, but a correct exposure level is required for the final quality of the exam," Primo said. "This is important for patient radiation protection."

MR. PAGE is a freelance writer in Redondo Beach, CA.


© 2001 CMP Media LLC, a United Business Media company