Four international trials involving full-field digital mammography (FFDM) offered recent updates on the modality, lauding it as a user-friendly method of detecting and following up breast cancer.
First at the 2003 RSNA meeting in Chicago was a Norwegian researcher who discussed the end results, including long-term and follow-up, of women enrolled in the Oslo I Screening Project, as well as the results of an observer performance study for breast lesion detection among the project’s mammographers.
Radiologist performance also was assessed in two related studies out of Germany and the Netherlands. This time, the investigators looked at whether training readers in soft-copy FFDM interpretation brought about better results.
The Oslo experience
Dr. Per Skaane from Ulleval University Hospital in Oslo, Norway shared the final results of his group’s screening project, which involved over 3,000 women. They found no statistically significant difference in the cancer detection rate between SFM and FFDM.
The goal of the population-based study, begun in January 2000, was to compare the cancer detection rate of screen-film mammography (SFM) to FFDM with soft-copy reading. The study also looked at interval cancers and screen-detection cancers in subsequent screening rounds.
In this paired study, 3,683 women underwent SFM and FFDM; two standard views of each breast were acquired on both exams. Independent double reading using a five-point rating scale of cancer probability was applied, with an exam defined as positive if at least one reader scored it as two or higher. Positive cases on both modalities were discussed at separate consensus meetings.
"The women included in the study were followed for two years with respect to interval cancers and screen-detected cancers in the subsequent screening round in order to reveal any differences in the cancer detection rate between the two modalities," Skaane and colleagues wrote.
According to the results, a total of 31 cancers (0.84% detection rate) were found during the initial interpretation sessions. Twenty-eight were found with SMF and 23 with FFDM.
Two cancers were dismissed at the SFM consensus meeting, and three at the FFDM meeting. A total of 10 interval cancers occurred during the two-year follow-up period. Six of these cancers had originally been assigned normal scores on SFM and FFDM. The remaining four cancers had positive interpretations (three on SFM, one on FFDM), but were dismissed at the consensus meetings. A total of 15 screen-detected cancers were found in the subsequent screening round two years later, of which 11 had a normal rating at initial interpretation.
Overall, 39 cancers were scored as positive on one or both modalities, with SFM detecting 32 and FFDM finding 28.
Skaane attributed the missed cancers to improper postprocessing of the images rather than observer error. However, new workstation designs, computer-aided detection (CAD), and reader training should boost cancer detection rates on FFDM.
In a second study, Skaane and co-investigators conducted an observer performance study for SFM (Mammomat, Siemens Medical Solutions, Erlangen, Germany) versus FFDM (Senographe 2000D, GE Medical Systems, Waukesha, WI) in the detection and classification of breast lesions. A total of 232 cases were displayed, including 46 cancers, 88 benign lesions, and 98 normal cases, for interpretation by six readers experienced in SFM and soft-copy digital reading.
Based on the results, one reader performed better with SFM while five did better with FFDM. Using a multireader analysis, the mean Az for all readers was 0.916 for FFDM and 0.893 for SFM (p value = 0.25).
For the cancer cases, FFDM resulted in the correct classification (BIRADS 3 or higher) of three additional cancers per reader.
Skaane concluded that while FFDM did allow for the correct classification of more breast cancers than SFM, the difference was not statistically significant. However, FFDM did provide at least similar lesion classification to SFM.
SCREEN-TRIAL project
Biologist Nora Wedekind of MeVis BreastCare in Bremen, Germany presented initial results from the ongoing European SCREEN-TRIAL. Specifically, the investigation focused on whether education can improve radiologists’ performance in soft-copy reading of FFDM exams. The images in this trial were all displayed on a MeVis BreastCare soft-copy reading station.
Some of the sites involved in the trial are the University Medical Center in Nijmegen, the Netherlands; the Bremen Breast Cancer Screening Program; and the Centre for Oncological Study and Prevention in Tuscany, Italy. MeVis, Siemens Medical Solutions of Malvern, PA, and R2 Technology of Los Altos are supporting the trial.
In the three-phase study, a dozen radiologists from five different sites participated.
All were experienced mammographers, although unfamiliar in soft-copy reading. In phase one, the readers received training in soft-copy system operation and reading. This was followed by a skills test, in which 150 FFDM cases (120 normal, 30 cancers) were read in both hard-copy and soft-copy formats.
In the upcoming phase two, all participants will read at least 1,000 soft-copy FFDM cases (with approximately 5% of the cases being cancerous) over a six-month period. Finally, the reader will reevaluate the original 150 cases in soft-copy format only.
In phase one, the reading speed was lower for soft-copy and the performance for detecting microcalcifications was slightly lower, Wedekind reported. However, after the mini training session, the radiologists were still able to read their cases in the pre-defined time of 75 cases in two hours. Wedekind said that once the 1,000-case self-training period is over, they anticipate improved results.
One of the sites in the Netherlands trial also reported its individual results at the RSNA conference. The caseload consisted of 150 two-view digital mammograms. There was a two- to four-week gap between the first read and the second, with the time limited to 75 cases in two hours. Digital images were acquired on a DMR 2000D with post-processing.
The training included operation and reading skills as well as a post-training assessment. Ton Roelofs, Ph.D, from the University Medical Center in Nijmegen said that for his site, lower results were posted for microcalcifications as well. Once again the 10 radiologists in this study posted a lower reading time with soft-copy, but that result was not statistically significant.
Overall, the results of phase one indicate that asking radiologists to do soft-copy reading without training is less than beneficial. The key to improving soft-copy reading will be more extensive education, Roelofs said. In addition, technical issues, such as a potential degradation of microcalcifications on soft-copy images, need to be addressed in order to study observer performance accurately.
By Shalmali PalAuntMinnie.com staff writer
January 16, 2004
Related Reading
Frost predicts tripling of non-mammo breast imaging in U.S., December 9, 2003
Investigators find x-ray dose reduction varies widely among digital mammo machines, December 3, 2003
Detours mark the road to U.S. adoption of digital mammo, November 21, 2003
Women’s imaging vendors target flexibility, CAD enhancements, November 17, 2003
FFDM fares as well as screen-film for population-based breast screening, December 1, 2002
With experience, readers prefer full-field digital mammo, November 26, 2001
Copyright © 2004 AuntMinnie.com
