The utilization of automated testing tools and image analysis software in Belgium and Luxembourg has allowed these countries to succesfully implement a centralized daily quality assurance program for all its CR- and DR-based digital mammography screening modalities with a minimal impact on network bandwidth and clinical personnel.
Files sizes of typical uncompressed digital mammography images range from 8 MB for an 18-cm x 23-cm field of view (FOV) at 100 μm, and can be up to 50 MB for a 24-cm x 30-cm FOV at 50 μm; lossless compression can reduce this by one-half to one-third, still presenting a significant hit to network bandwidth when multiple systems are tested on a daily basis.
Belgium has 46 digital mammography units and Luxembourg has two that are required to comply with quality control (QC) directives as specified in the European Guidelines on Digital Mammography (EUREF), said Hilde Bosmans, who presented an overview of the digital mammography quality initiative at the 2008 International Workshop on Digital Mammography earlier this summer in Tucson, Ariz.
Bosmans, from the radiology department at Leuven University in Leuven, Belgium, reported that CR systems from Agfa Healthcare, Fujifilm Medical Systems and Konica Minolta Medical Imaging are deployed in the countries as are DR systems from GE Healthcare, Hologic, IMS and Siemens Medical Solutions.
“In a decentralized screening environment with centralized quality assurance, QC supervision can become time-consuming work,” she noted.
On a daily basis, a technologist at every facility takes two exposures of a 40-mm polymethyl methacrylate (PMMA) tissue-equivalent material plate on each digital mammography system. The raw images are sent automatically to locally-installed analysis software, which extracts specified QC data and then calculates a 2-cm x 2-cm region of interest (ROI) over the complete image.
“To save network bandwidth, only the analysis results and ROI data are sent to reference site,” Bosmans said.
At the centralized reference site, 3D plots, color contour plots, deviation maps and thumbnail images are created from the analysis and ROI results, which are then examined and compared with prior results by a trained physicist.
“Several artifacts can be traced with the thumbnail images,” she said. “These include dirt on phosphor cassettes, scan line artifacts, scratches on the imaging plate and burned in markers for the CR units. For DR units, reduced sensitivity of detector material, lag images, crystallization of detector material, defective pixels, faulty detector columns, bad calibration artifacts, blooming and several electrical artifacts were noticed.”
Over a year-long system deployment, the team has fine-tuned its QC parameters to eliminate redundant data, enabling it to reduce the file size for evaluation to below 1 MB, Bosmans noted. This has allowed the group to simplify remote QC for mobile screening units and centers with limited network infrastructure or network access.
“In the constancy check procedure proposed by the EUREF, it is stated that QC should be done on a weekly basis,” she said. “We increased the frequency from a weekly to a daily test in the Flemish breast cancer screening program. This enables us to find problems sooner and also to have a better understanding of a system’s performance over time.”
Although the system was developed for mammography QC, with minor modifications-- such as changing the position of the reference ROI’s--it could be used to test the homogeneity of DR and CR systems for general radiography, Bosmans concluded.