The emergence of a new, powerful imaging modality is cause for both celebration and consternation, and digital breast tomosynthesis (DBT) has proven no exception to this rule, according to participants in a June 8 educational forum at the 2013 meeting of the Society for Imaging Informatics in Medicine (SIIM), held in Grapevine, Texas. Early results from sites offering DBT to their patients have been nothing short of extraordinary: X-ray Associates of New Mexico (XRANM) in Albuquerque, for instance, reports a 48% reduction in its recall rate, while the University of Pittsburgh Medical Center (UPMC) in Pennsylvania has seen a 40% increase in detection of invasive breast cancers, with a reduction in false positives of 15%.
The journey to achieving these exciting results with a sustainable workflow is not without its challenges, however. In the session “Digital Breast Tomosynthesis and the Informatics Infrastructure: How DBT Kills Your PACS/VNA,” a broad group of 24 physicians, informaticists, and vendors met to discuss the challenges of incorporating the new modality into both radiologist workflow and existing informatics infrastructure.
Michael Linver, MD, of XRANM, observes that his facility’s recall rate has plummeted since the implementation of DBT. “Our recall rate is down 48% for patients who had DBT (versus those who did not), which cuts way back on the number of patients we’re having to see on the diagnostic side,” he observes. “The problem, for us, was interpretation. DBT can take five times as long to read as a regular screening mammogram, and that was a bit of a hardship. It makes our days, as radiologists, even longer.”
With time of the essence, alternating between a standard PACS workstation and the proprietary Hologic workstation is not a viable option, for many facilities. The issue, as outlined by speaker Donna Plecha, MD, a radiologist with University Hospitals Case Medical Center (UHCMC), Cleveland, Ohio, is that the data issued by the Hologic breast-tomosynthesis system are in a proprietary format. When her organization began reading DBT studies, in September 2011, “Our PACS couldn’t accept the tomosynthesis images, so we’d read 2D mammograms on our Sectra PACS, get out of our chairs, sign in at the Hologic workstation to read the 3D images, and do a screen capture of one marked image to send to our PACS,” Plecha says.
Speaker Margarita Zuley, MD, director of breast imaging at UPMC, reports similar problems. She outlines the key challenges related to dealing with tomosynthesis images: sizing, outside image management, orientation, procedure codes, and the creation of synthetic 2D images from 3D datasets. “Some PACS can store DBT images but cannot display them, and some can’t even store them,” she notes. Even for PACS platforms that can view and store DBT images, projection data remain proprietary to Hologic, she adds.
Now, however, vendors are rushing to make DBT images compatible with their applications. One of the first to achieve this goal was Sectra; by collaborating with Hologic, the company was able to integrate DBT images into its breast imaging platform, enabling physicians to read them side by side with more traditional 2D screening mammograms. “We’ve been able to read screening DBT images on our Sectra PACS since August 2012,” Plecha says. “We’re able to make hanging protocols on the 2Ds that we can compare with the 3Ds, as well as with prior exams.”
Stamatia Destounis, MD, attending radiologist at Elizabeth Wende Breast Care (EWBC), Rochester, New York, reports a similar challenge and solution. “Having to read on the Hologic workstation was a real workflow breakdown,” she says. “Sectra was able to allow us to view the 3D images on our mammography PACS and look at them side by side with prior exams and prior 2D mammograms. Reading the combined images probably still takes twice as long, but our prior workflow, with the Hologic workstation, really limited how many DBTs we could do. Having them integrated into the PACS represents a real clinical advantage. This is the way it should be done.”
All three facilities have seen inspiring results from offering DBT, making the extra work well worth it. “We’ve found a close to 50% reduction in recall rates with combined 2D and 3D screening, and we are identifying more cancers, including small masses, smoothly outlined masses near the nipple, and calcifications,” Destounis says. Plecha adds, “It’s challenging, but it’s worth it. When you see your cancer-detection rates go up, it’s worth the extra time it takes to read. We haven’t made this big a jump in breast imaging since we started with digital mammography.” Linver concludes, “It’s made us better at what we do, and that’s the bottom line.”
Behind the scenes, PACS administrators and informaticists at these organizations have been hard at work adapting IT infrastructure to handle images that can be 20 times the size of datasets for full-field digital mammography—500MB, versus 25MB for a screening mammogram. “The radiologists were experiencing some slowness loading the images because they were such large files,” Beverly Rosipko, PACS administrator for UHCMC, explains. “We had to have the memory upgraded on our workstations; now, we have up to 24GB of RAM, whereas standard workstations only have 8GB.”
Diana Kissel, PACS administrator for EWBC, notes that the organization had already upgraded its workstations prior to DBT implementation—but the images’ file sizes can cause longer-term problems as well. “These images are 10 to 20 times larger than regular 2D mammograms, meaning we will need to increase our storage space,” she says. Sagit Frasier, COO of XRANM, found it necessary to keep DBT images in-house until they could be read by radiologists because of the amount of time that it takes to fetch them from the remote PACS. She also advises fellow informaticists to invest time in hanging protocols. “You want to make everything as easy as possible for your physicians because they’re already dealing with this additional study volume,” she says. “Your existing hanging protocols will not work, so you want to build these beautifully, and pay attention to hanging for DBT prior studies as well.”
Zuley also notes that DBT images are transmitted between the modality and the PACS as DICOM breast-tomosynthesis objects, using a new format that doesn’t include projection images. Frasier notes, however, that in her experience, projection images are rarely needed by radiologists. “Those images can be important when the radiologist suspects motion,” she says. “Our physicians called around to other organizations to hear about their experiences, and what they heard was that if they had the ability to read everything on the Sectra PACS, versus having to read the DBTs on the Hologic workstation, the PACS workflow would be preferable. Within a few days, we realized how rarely the projection images were needed—it happens less than five times a year.”
Participants in the SIIM forum report similar challenges, and they call for the development of an Integrating the Healthcare Enterprise profile extension to help solve the problem, going forward. Panelists also recommend that organizations implementing DBT upgrade their networks to at least 1GB and archive images as database objects from the outset (to avoid migration problems later). Panelist Herman Oosterwijk, PACS consultant and president of OTech, says, “Translating this innovation into clinical practice is going to be disruptive.” In spite of the hurdles, however, Rosipko concludes, “It’s exciting to be involved in something so new and so great for women.”