EMBARGOED FOR RELEASE: 31 AUGUST 2000 AT 17:00 ET US
Contact: David Williamson
rdtokids@email.unc.edu
919-962-8596
University of North Carolina at Chapel HillResearch shows radiologists will need different programs for digital mammography to work best
CHAPEL HILL - For doctors to take best advantage of new digital mammography and possibly save more women from breast cancer, companies that manufacture the technology need to create new, more sophisticated computer programs, University of North Carolina at Chapel Hill scientists say.
That's because a team of researchers led by Drs. Etta Pisano, professor of radiology at the UNC-CH School of Medicine, has found that radiologists prefer different kinds of images for evaluating different types of breast tissue abnormalities. "Before we did our study, we thought that a certain type of lesion and a certain kind of breast density, for example, would require only one type of processing," said Pisano, also chief of breast imaging at UNC-CH and a member of the UNC Lineberger Comprehensive Cancer Center. "Now, we think displaying images will be more complicated. To get full use out of digital mammography machines, companies are going to have to put some resources into optimizing images for different tasks."
An image processing program that reveals calcifications in the breast will be different from one that helps doctors evaluate those calcifications once they've been found, she said. The same holds true for cancerous breast tumors.
A report on the latest research appears in the September issue of Radiology, a top medical journal. Besides Pisano, authors are Elodia B. Cole, research associate in radiology; Stacey Major, former research associate in biostatistics; Shuquan Zong, graduate research associate in biomedical engineering; Bradley M. Hemminger, senior research associate in radiology; Dr. Keith E. Muller, associate professor of biostatistics; and 27 co-authors from the International Digital Mammography Development Group. The latter is a consortium of radiologists from UNC-CH, Harvard, Johns Hopkins, Duke and other universities. A companion report in the September issue of the journal Radiographics contains images from the study.
The research involved showing images from 28 patients with various known breast abnormalities to 12 radiologists who compared digital images from three different commercially available machines to traditional film mammograms from the same patients. The physicians were asked to decide and record which images they preferred in making diagnoses. Manufacturers were General Electric, Fisher and Trex.
"When digital mammograms were preferred over film mammograms, the radiologists in our group selected different forms of digital image processing for each of three different mammography reading tasks and for different breast lesions," Pisano said. "That means that if we make all the images look like regular mammograms, we will be limiting their effectiveness."
The challenge for General Electric and the other companies will be to create new, more sophisticated computer software that takes into account what radiologists believe provides the most helpful pictures, the physician said. Digital mammography holds great promise, she said. "We believe that it will eventually prove to be a better tool than the currently available technology, film mammography, in finding breast cancer earlier when it is still curable," Pisano said. "But unless we take the time and money to figure out how the images should be displayed for different tasks, this technology might not meet our high expectations, and the effect on women's health might not be as big as it could be."
### The National Cancer Institute, the U.S. Army Medical Research and Material Command and the Canadian Breast Cancer Research Initiative supported the research.
Note: Pisano can best be reached at etpisano@med.unc.edu. Her number is (919) 966-6957.
Contact: David Williamson, (919) 962-8596.
Back to EurekAlert!
Journal
Radiology