Early experience in the use of quantitative image quality measurements for the quality assurance of full field digital mammography x-ray systems

Abstract

Quantitative image quality results in the form of the modulation transfer function (MTF), normalized noise power spectrum (NNPS) and detective quantum efficiency (DQE) are presented for nine full field digital mammography (FFDM) systems. These parameters are routinely measured as part of the quality assurance (QA) programme for the seven FFDM units covered by our centre. Just one additional image is required compared to the standard FFDM protocol; this is the image of an edge, from which the MTF is calculated. A variance image is formed from one of the flood images used to measure the detector response and this provides useful information on the condition of the detector with respect to artefacts. Finally, the NNPS is calculated from the flood image acquired at a target detector air kerma (DAK) of 100 µGy. DQE is then estimated from these data; however, no correction is currently made for effects of detector cover transmission on DQE. The coefficient of variation (cov) of the 50% point of the MTF for five successive MTF results was 1%, while the cov for the 50% MTF point for an a-Se system over a period of 17 months was approximately 3%. For four a-Se based systems, the cov for the NNPS at 1 mm−1 for a target DAK of 100 µGy was approximately 4%; the same result was found for four CsI based FFDM units. With regard to the stability of NNPS over time, the cov for four NNPS results acquired over a period of 12 months was also approximately 4%. The effect of acquisition geometry on NNPS was also assessed for a CsI based system. NNPS data acquired with the antiscatter grid in place showed increased noise at low spatial frequency; this effect was more severe as DAK increased. DQE results for the three detector types (a-Se, CsI and CR) are presented as a function of DAK. Some reduction in DQE was found for both the a-Se and CsI based systems at a target DAK of 12.5 µGy when compared to DQE data acquired at 100 µGy. For the CsI based systems, DQE at 1 mm−1 fell from 0.49 at 100 µGy to 0.38 at 12.5 µGy. For the a-Se units, there was a slightly greater reduction in average DQE at 1 mm−1, from 0.53 at 100 µGy to 0.31 at 12.5 µGy. Somewhat different behaviour was seen for the CR unit; DQE (at 1 mm−1) increased from 0.40 at 100 µGy to 0.49 at 12.5 µGy; however, DQE fell to 0.30 at 420 µGy. DQE stability over time was assessed using the cov of DQE at 1 mm−1 and a target DAK of 100 µGy; the cov for data acquired over a period of 17 months for an a-Se system was approximately 7%. For comparison with conventional testing methods, the cov was calculated for contrast-detail (cd) data acquired over the same period of time for this unit. The cov for the threshold contrast results (averaged for disc diameters between 0.1 mm and 2 mm) was 6%, indicating similar stability.