PACS algorithms promise improved efficiency for radiologists

Soft-copy workstations provide the ability to manipulate and display images-but few radiologists take advantage of it. Instead they opt for traditional views that mimic those of film and light boxes. But a study has found that by incorporating image-processing algorithms that highlight specific types of pathology into workstations, radiologists render potentially faster and more accurate diagnoses.

In the study, Dr. Bruce Reiner, director of radiology research at the University of Maryland Health System in Baltimore and colleague Dr. Eliot Siegel found that the interpretation of soft-copy chest images was most accurate when using the combined processing techniques of gray-scale inversion and edge enhancement. The researchers based their conclusions on a study of 100 chest images acquired using a Fuji CR system and enhanced with tools available on the Fuji Synapse PACS workstation. Processing formats included standard default, edge enhancement, dynamic range control, and gray-scale inversion with edge enhancement.

“The idea was to change the way we traditionally do things in the film environment in order to complement the complete tool kit of the computer workstation,” Reiner said. “Our hypothesis was that we could modify radiology images to maximize our interpretation accuracy without sacrificing productivity.”

As radiology departments change over to filmless operations, it will become more important to make the most of the benefits of PACS and related image processing capabilities, not only for efficiency but also for clinical benefit, he said. To do so, radiology needs to break with the past.

Radiologists traditionally work with a single two-view chest study. Computers offer the possibility of unlimited permutations.

“But most radiologists still read in that traditional paradigm, in which they look at the image in the single best presentation state,” Reiner said. “What we have found is that it might be preferable to present the image in multiple states: algorithms that highlight bony structures, the air-containing space, the lung, and the interstitial. Multiple presentations might be the best way to look at different types of pathologies.”

Other digital modalities already make use of this concept. CT scans of the lung, for example, typically include preset lung window and soft-tissue views. Similar shortcuts might be built into soft-copy chest images to look for pulmonary nodules or edema, Reiner said.

But there may be a trade-off. Viewing multiple images takes more time than the traditional “best view” approach to interpretation.

“Intuitively, if you are looking at three or four images instead of just one, yes, it is going to take more time,” Reiner said. “That’s where creativity comes in: we need to incorporate these algorithms into the functionality and design of the workstation.”

In the future, computer-aided detection and artificial intelligence programs may be built into workstations, he said. With these tools, radiologists might be able to identify pathology more quickly and accurately.

“Radiologists are trained today in a film environment and then have to apply those patterns to the workstation environment,” he said. “What we are talking about is changing the way radiologists practice in a way that improves efficiency.”

Reiner and Siegel are looking for feedback from radiologists. An interactive exhibit at the May Society for Computer Applications in Radiology meeting yielded some. More may come from a planned display at the upcoming RSNA meeting. Ultimately, Reiner hopes to take the experiment to the Internet, where Fuji CR users from around the world can assist in designing disease-specific algorithms.