Signal uniformity of mammography systems and its impact on test results from contrast detail phantoms

Technical quality assurance (TQA) procedures for mammography systems usually include tests with a contrast-detail phantom. These phantoms contain multiple objects of varying dimensions arranged on a flat body. Exposures of the phantom are then evaluated by an observer, either human or software. One well-known issue of this method is that dose distribution is not uniform across the image area of any mammography system, mainly due to the heel effect. The purpose of this work is to investigate to what extent image quality differs across the detector plane. We analyze a total of 320 homogeneous mammography exposures from 32 radiology institutes. Systems of different models and manufacturers, both computed radiography (CR) and direct radiography (DR) are included. All images were taken from field installations operated within the nationwide Austrian mammography screening program, which includes mandatory continuous TQA. We calculate signal-to-noise ratios (SNR) for 15 regions of interest arranged to cover the area of the phantom. We define the 'signal range' of an image and compare this value categorized by technologies. We found the deviations of SNR greater in anterior-posterior than in lateral direction. SNR ranges are significantly higher for CR systems than for DR systems.