Digital mammography: DQE versus optimized image quality in clinical environment an on site study

Transcription

1 Digital mammography: DQE versus optimized image quality in clinical environment an on site study Nadia Oberhofer* a, Alessandro Fracchetti a, Margareth Springeth a, Ehrenfried Moroder a a Health Service South Tyrol, Radiation Physics Dep., Regional Hospital, Bolzano, Italy ABSTRACT The intrinsic quality of the detection system of 7 different digital mammography units (5 direct radiography DR; 2 computed radiography CR), expressed by DQE, has been compared with their image quality/dose performances in clinical use. DQE measurements followed IEC using a tungsten test object for MTF determination. For image quality assessment two different methods have been applied: 1) measurement of contrast to noise ratio (CNR) according to the European guidelines and 2) contrast-detail (CD) evaluation. The latter was carried out with the phantom CDMAM ver. 3.4 and the commercial software CDMAM Analyser ver. 1.1 (both Artinis) for automated image analysis. The overall image quality index proposed by the software has been validated. Correspondence between the two methods has been shown figuring out a linear correlation between CNR and. All systems were optimized with respect to image quality and average glandular dose (AGD) within the constraints of automatic exposure control (AEC). For each equipment, a good image quality level was defined by means of CD analysis, and the corresponding CNR value considered as target value. The goal was to achieve for different PMMA-phantom thicknesses constant image quality, that means the CNR target value, at minimum dose. All DR systems exhibited higher DQE and significantly better image quality compared to CR systems. Generally switching, where available, to a target/filter combination with an x-ray spectrum of higher mean energy permitted dose savings at equal image quality. However, several systems did not allow to modify the AEC in order to apply optimal radiographic technique in clinical use. The best ratio image quality/dose was achieved by a unit with a-se detector and W anode only recently available on the market. Keywords: Digital mammography, image quality, DQE, a-se direct conversion detector 1. INTRODUCTION In mammography image quality requirements are very demanding. Particularly in relation to mammography screening it is important to achieve very high image quality with the lowest possible dose. A large improvement in this task has been realized by the introduction of digital image receptors, combined with monitors for softcopy image display. This allows to address the optimization of image acquisition separately from image visualization. The potential of so called direct radiography (DR) systems for considerable dose reductions at equal image quality is generally recognized 1,2,3. Computed radiography (CR) systems adopting photostimulable storage phosphor plates as radiation detection device may allow progress, too, but on a minor scale. However, on site measurements show that the mere introduction of digital technology is not equivalent to dose saving. In the daily routine, the switch from an analog or elder digital technology to a new digital system often has to occur in the shortest possible time interval. In particular in the case of CR units, where the radiological equipment does not change, exposure settings are sometimes just adapted from the previous anlog modality, without any critical review. On the acquisition stations and workstations monitors all images are displayed with a similar gray level, regardless of the radiation amount used for image production. Hence in digital systems the possibility of checking correct patient exposure straightforward, as was feasible just glancing at film blackening, does not exist any more. Overexposures are not recognizable by the displayed image. Furthermore the common materials used as radiation converters in both types of digital image receptors (DR and CR) exhibit different energy absorption characteristics than the converter screens in an analog system, requiring a revision of the energy matching between x-ray tube output and detector. This means that accurate exposure settings optimization with respect to image quality and patient dose is an even more pressing challenge within digital technology. *nadia.oberhofer@asbz.it; phone 0039 (0) ; fax 0039 (0) Medical Imaging 2010: Physics of Medical Imaging, edited by Ehsan Samei, Norbert J. Pelc, Proc. of SPIE Vol. 7622, 76220J 2010 SPIE CCC code: /10/$18 doi: / Proc. of SPIE Vol J-1

2 Detective quantum efficiency (DQE) determination is generally regarded as the gold standard for characterizing the intrinsic quality of an imaging system. But the knowledge of the physical characteristics of a system does not allow to infer anything about image quality nor delivered dose in its clinical use. In fact DQE measurements, as standardized by IEC , are preferably carried out with specific radiation qualities, in mammography usually RQA-M2 (Mo target, 0.03 mm Mo filtration, 28 kv), whereas patient exposures may be executed with very different ones. The radiation protection entity of interest in mammography, the average glandular dose (AGD), depends strongly on radiation quality. As known, the use of target/filter combinations corresponding to x-ray spectra of higher mean energy with respect to Mo/Mo may allow dose savings for the patient 4,5,6,7. In order to assess and optimize the performance of a radiographic system it is necessary to quantify image quality. Two common approaches are 1) measurement of contrast to noise ratio (CNR) and 2) contrast detail (CD) evaluation. The first relies on the fact that in radiographic images signal perceptivity by human observers is contrast driven 8. CNR is an objectively measurable parameter. Unfortunately its value is a relative quantity, which has significance only if compared to values of the same system. The simplicity of its measurement, however, makes it the ideal working parameter for equipment optimization 9. To implement the second method, contrast detail evaluation, several phantoms have been proposed. They all feature a series of small details with different diameters and background contrasts. The identification of the lowest detectable contrast for each diameter, i.e. the threshold contrast, gives a contrast-detail-curve. If contrast detail images are analyzed by human observers, the problem of inter- and intra-observer variability arises. This can be overcome by multiple evaluation or by automated image analysis, which provides reproducible, observer independent results 10,11. Under this conditions CD evaluation can be considered as an absolute measurement of image quality. In fact it is used in the European Guidelines 12 to define two levels of image quality: a lower acceptable level and a higher achievable one. Thus CD evaluation is appropriate for comparison between different systems. Data analysis may be facilitated if the threshold curve behavior is summarized by a single performance figure 13. In this work we investigate for several mammography equipments how detection performance, expressed in terms of DQE, traduces into image quality/agd ratio in clinical use. 2. MATERIALS AND METHODS 7 Systems with 6 different detector or radiation production technologies have been studied. Their basic properties are summarized in Tab. 1. Systems A1 and A2 are twin systems. All 5 DR systems use as detector an active matrix flat panel imager (AMFPI) with linear response to dose. The two digitizers of the CR systems C1 and C2 use the same reading technology and differ just in throughput. Their response curve follows a square root law. Images were linearized before DQE evaluation. All analyzed images were dicom images for processing. CR images were processed with the option System Diagnosis, Flat Field Mammo, which applies only corrections for pixel gain equalization. Dose measurements were taken in terms of kerma air with a regularly calibrated ionization chamber (Radcal X6-6M, Radcal Corporation, USA). AGD was calculated as described in the European Guidelines 12. Table 1. Basic properties of studied digital mammography systems System Model (Manufacturer) Target/Filter in clinical use Detector type (Manufacturer), pixel dimensions A1, A2 (DR) Selenia (Lorad) Mo/Mo, Mo/Rh direct amorphous Se (Hologic), 70 µm A3 (DR) Mammomat Inspiration (Siemens) W/Rh direct amorphous Se (Anrad), 85 µm B1 (DR) Senographe DS (GE) Mo/Mo, Mo/Rh, Rh/Rh indirect CsI (GE), 100 µm B2 (DR) Senographe 2000D (GE) Mo/Mo, Mo/Rh, Rh/Rh indirect CsI (GE), 100 µm C1 (CR) CR85/DMR+ (Agfa/GE) Mo/Mo, Mo/Rh, Rh/Rh Powder phosphor (AGFA), 50 µm C2 (CR) CR35/Sophie (Agfa/Planmed) Mo/Mo Powder phosphor (AGFA), 50 µm Proc. of SPIE Vol J-2

3 2.1 MTF, NNPS and DQE evaluation DQE measurements followed IEC using for presampled modulation transfer function (MTF) determination a tungsten test object placed either directly on the CR plate or on the Bucky in DR systems. Exposures were made with radiation quality RQA-M2 (Mo target, 0.03 mm Mo filtration, 28 kv, 2 mm Al additional filtration) with neither grid nor compression device in place and manual load choice. Images for MTF determination were acquired with a detector entrance air kerma corresponding to the values used by each unit in clinical automatic exposure control (AEC) modality for a 4.5 cm polimethylmetacrylate (PMMA) block (from 72 μgy to 99 μgy), which we will denote as normal exposure level. Evaluation was performed with the open source Java plug in Flat-panel for ImageJ, extended for DQE calculation. To reduce the statistical noise, the reported normalized noise power spectrum (NNPS) is a mean value derived from 3 images (successive exposures at identical exposure settings). Correctness of calculation routine has been verified previously by comparison of results with literature, when possible14, Image quality assessment Two different methods have been applied: measurement of contrast to noise ratio (CNR) and contrast detail (CD) evaluation. CNR was determined according to the European Guidelines12. The test object consisted of several PMMA layers covering the whole detector area with a contrast object (4 cm2 foil of 0.2 mm Al) positioned on top, 6 cm from the chest wall side. Total phantom thickness varied from 2 cm to 7 cm. CNR was calculated considering two regions of interest (ROIs) of approximate 1 cm2 area each on the image, one inside the contrast object (Al), the other next to it on the background, as indicated in Figure 1 (left). The adopted formula was CNR = MPVAl MPVbackground 2 SDAl2 SDbackground + 2 2, (1) where MPV = mean pixel value SD = standard deviation. Error of the CNR value was estimated as standard deviation of the evaluation of 10 repeated exposures. Bkg For CD evaluation the phantom CDMAM ver. 3.4 (Artinis, The Netherlands) was used. Figure 1 (center) shows a radiography of the detail tablet (18 cm x 24 cm x 0.5 cm). The contrast objects are accomplished as thin gold discs of exact diameter and thickness, arranged in squares. Within a column the disc-diameter is constant, with exponentially increasing thickness, and within a raw the disc thickness is constant, with exponentially increasing diameter. Each square contains two details, one in the center and one in a randomly chosen corner. The task consists in the correct detection of the corner detail position. The detail plate was always positioned above 2 cm of PMMA with the smaller details towards the chest wall side. Its absorption corresponds to 1 cm PMMA. Additional PMMA plates needed to achieve the desired total phantom thickness were put on top of the detail plate towards the tube. Al 1 Figure 1. Left: Experimental setup for CNR measurement. A 0.2 mm Al foil is used al contrast object. Center: Phantom CDMAM 3.4 (Artinis, Nijmegen, The Netherlands) for contrast detail evaluation. Right: Output example of the commercial software CDMAM Analyser 1.1 (Artinis) for automated CD evaluation. Proc. of SPIE Vol J-3

4 CD analysis was carried out with the commercial software CDMAM Analyser ver. 1.1 (also Artinis) for automated image analysis. In order to reduce uncertainty, all reported results represent the curve fitted value of the analysis of a set 8 raw images (detection rate of 75% throughout the study). Figure 1 (right) shows an example of software output. The darkness of the dots is related to the detection frequency of the corner details, being highest if one detail has been correctly detected in all images. In DR systems the phantom was slightly moved between successive images. Results are expressed in terms of an overall image quality index, the Image Quality Figure Inverse, defined as 16 IQF 100 inv = Threshold, (2) i= 1 Diameteri * where Threshold i is the thickness of the thinnest visible detail with diameter i. increases when the number of correctly detected details growths. It has previously been demonstrated that this index represents an objective and absolute measure of the image quality and is suitable for comparison of different equipments 13, if the same phantom is used. The correlation between CNR and has been studied. This required to evaluate 1 CNR image and a series of 8 CDMAM images for each experimental setting (kv, mas, target/filter, phantom thickness). For all DR systems a very wide range of exposure combinations has been investigated, including unusual ones. As long as the image noise is dominated by quantum noise, applying a higher dose improves image quality. It is considered CNR 2 dose. Relying on the linear relation between CNR and, the figure of merit IQF 2 inv /AGD was defined to assess the influence of target/filter combination on image quality and dose. 2.3 Mammography system optimization Before evaluation, the mammography systems were optimized with respect to image quality and AGD within the constraints of AEC devices. The goal was to achieve for different PMMA-phantom thicknesses (2 cm 7 cm) constant image quality, that means a constant CNR value, at minimum dose 9. On every equipment first a set of CDMAM images at normal exposure conditions was taken and CD evaluation performed, to check compliance with the requirements for achievable image quality reported in the European Guidelines. Otherwise a second image set was taken with higher mas value. The calculated was recorded for image quality comparison between different units. Keeping the same settings, an exposure for CNR determination was made. The resulting CNR value was considered as specific target value for the unit under investigation. Then phantom thickness was changed and CNR calculations performed. The exposure parameters (target, filter, kv, mas) were varied in order to obtain the target CNR value at the lowest possible AGD. The parameter CNR rel [%], which normalizes CNR values at different phantom thicknesses to the CNR value at the reference thickness of 5 cm, has been used to compare with limits of the European guidelines. After optimization the absolute image quality of the different systems, described by, was compared at various AGD levels. 3. RESULTS 3.1 MTF, NNPS and DQE evaluation The parameters were determined in two directions: parallel and orthogonal to the chest wall side. For System B1 only the MTF data are available. All DR systems exhibited a considerable higher MTF compared to CR systems (see Figure 2 left). As expected, systems with direct conversion detectors (a-se) showed higher MTF values than systems using indirect conversion detectors (CsI). Systems A1, A2, B1 and B2 did not show any significant direction dependence in the calculated MTF value. In System A3 the calculated MTF resulted lower if evaluated parallel to the chest wall side (-20% at f = 5 mm -1 ), in the CR systems C1 and C2 it was the scan direction to have inferior MTF. The right part of Figure 2 summarizes the findings for NNPS at normal exposure levels (from 72 μgy to 99 μgy) along the two axis; Figure 4 illustrates 2-dimentional NNPS representations. Whilst MTF values, as known, are not dose sensitive, NNPS decreased in all systems when dose increased. CR systems and the indirect conversion detector of System B2 exhibited a reduction of NNPS with increasing frequency, whereas all direct conversion detectors produced almost constant noise contributions at all frequencies. The 2-dimensional visualization of the NNPS reported in Figure 4 unveiled particular noise components in some systems: bright spots arranged in a regular pattern in Systems A1 and A2 and bright, equidistant lines in Systems C1 and C2. These structures are visible in Figure 2 as marked peaks in the NNPS value calculated along the direction perpendicular to the chest wall. The CR systems exhibited the strongest frequency dependence, delivering the lowest NNPS values above 2 mm -1 and the highest above 1 mm -1. i Proc. of SPIE Vol J-4

5 NNPS [mm -1 ] 1,0E-04 1,0E-05 A1 A2 A3 B2 C1 C2 NNPS Hor (93,3 µgy) NNPS Hor (93 µgy) NNPS Hor (88 mas) NNPS Hor (90 µgy) NNPS Subscan (99,9 µgy) NNPS Subscan (72,4 µgy) NNPS Vert (93,3 µgy) NNPS Vert (93 µgy) NNPS Vert (88 µgy) NNPS Vert (90 µgy) NNPS Scan (99,9 µgy) NNPS Scan (72,4 µgy) A3 A2 1,0E-06 B2 A1 C2 1,0E Frequency [mm -1 ] Figure 2. Left: MTF of 7 digital mammography systems calculated parallel and perpendicularly to the chest wall side at ca. 90 μgy detector entrance kerma air. When there was no substantial difference, only one direction is reported. Right: Corresponding NNPS values. Vert / Scan direction = parallel to chest wall side. Error MTF: < 2%; error NNPS: < 5%. C1 Figure 3. DQE for 6 mammography systems calculated parallel and perpendicularly to the chest wall side at detector entrance kerma air of Figure 2 (ca. 90 µgy). When there was no substantial difference, only one direction is reported. Vert and Scan direction mean parallel to chest wall side. Error on DQE: ca. 8% of value. Figure 3 reports the results for DQE. The two CR systems brought about the lowest performance with peak values around 40% decreasing rapidly with frequency. The peak DQE of DR systems laid between 50% and 72%. In Systems A1 and A2 we found substantial drops in the DQE values along the direction perpendicular to the chest wall; apart from that the values generally differed only slightly with the evaluation direction. System B2 with the indirect conversion detector did not show any direction dependence. Its DQE results were slightly beneath those of the DR systems with direct conversion detector. System A3 produced the highest peak DQE value, but its values decreased more rapidly than in other DR systems with growing frequency. 3.2 Image quality assessment The two quantities CNR and employed for image quality assessment showed a strong correlation, which could be regarded as linear for all practical purposes, being the correlation coefficient r > 0.85 in all systems. Some examples related to DR systems are reported in Figure 5 a), b) and c). Figure 5 d), e) and f) show the outcomes of the figure of merit IQF 2 inv /AGD. Reported data represent a mean value of several points corresponding to AGD at clinical values and above. In all systems, use of the target/filter combinations which generate x-ray spectra of higher mean energy compared to Mo/Mo, that is Mo/Rh, W/Rh or Rh/Rh, resulted in a higher IQF 2 inv /AGD for almost every phantom thickness. For Proc. of SPIE Vol J-5

6 a) System A1+A2: AMFPI, a-se 70 µm b) System A3: AMFPI, a-se 85 µm c) System B1+B2: AMFPI, CsI 100µm d) System C1+C2: CR, 50 µm Figure 4. 2-dimensional representation of NNPS for different detector technologies. thickness 3 cm differences between similar radiation qualities, that means Mo/Rh and Mo/Mo (System A2) or Rh/Rh and Mo/Rh (System B1) were less pronounced. Figure 5 g) and h) compare for different phantom thicknesses (equivalent to 3, 5 and 6 cm PMMA) the systems capabilities, considering for each unit the optimum beam quality, i.e. the hardest one. At small phantom thickness (3 cm) and very low AGD System A2 outperformed the others, but only if Mo/Rh was used. At a medium phantom thickness of 5 cm differences between the systems were reduced: Systems A2 and A3 performed equally well. For a larger phantom thickness (6 cm) the W/Rh combination of system A3 seemed to produce a small advantage. For every thickness at clinical AGD System B1 performed slighly less. The difference has been checked to be significant at p=0.05 with the Wilcoxon-Test. According to the outcomes of DQE evaluation, at a given AGD CR systems scored significantly lower values than DR systems, as can be deduced from Figure 8, which compares the of the optimized systems for a medium phantom thickness and an AGD of 1.45 mgy. 3.3 Mammography system optimization Figure 7 reassumes the AGD levels for all systems at different phantom thicknesses (2 cm 7 cm) in exposures carried out in AEC modality after optimization. In the right part the corresponding relative CNR rel [%] values are depicted. Figure 6 reports two examples for system optimization. Proc. of SPIE Vol J-6

7 a) 2 /AGD [arb. units] 3,0 2,5 2,0 System A2 Mo/Rh Mo/Mo d) 1,5 1,0 0,5 0, Equiv. PMMA phantom thickness [cm] b) System A3 R 2 = 0, /AGD [arb. units] 3,0 2,5 2,0 1,5 System A3 W/Rh Mo/Mo e) cm MoMo 24 kv 2 cm WRh 30 kv 3 cm MoMo 26 kv 3 cm WRh 27 kv 5 cm MoMo 28 kv 5 cm WRh 29 kv 5 cm MoMo 29 kv 6 cm WRh 30 kv 7 cm WRh 31 kv Lineare CNR 1,0 0,5 0, Equiv. PMMAphantom thickness [cm] c) 2 /AGD [arb. units] 3,0 2,5 2,0 System B1 Rh/Rh Mo/Rh Mo/Mo f) 1,5 1,0 0,5 0, Equiv. PMMAphantom thickness [cm] g) 2 /AGD [arb. units] 3,0 2,5 2,0 System A2 System A3 System B1 h) 1,5 1,0 0,5 Figure 5. Data relative to Systems A2, A3 and B1. a), b), c): Examples for linear correlation between and CNR over a wide range of exposure parameter settings. d), e), f): Study of the figure of merit 2 /AGD at different PMMA thickness and beam qualities. Reported data represent a mean value of several points corresponding to AGD at clinical values and above. g), h): Comparison of outcomes related to the best target/filter combinations of each equipment. Error on : 5%, error on CNR: 3%, error on 2 /AGD: 10%. 0, Equiv. PMMA phantom thickness [cm] Proc. of SPIE Vol J-7

8 Figure 6. Example of optimization. Left: System B1. Right: System C2. Top: Trials for target CNR determination at different phantom thicknesses. Center: Calculated CNR rel [%] for exposures with AEC before and after optimization. Bottom: Corresponding AGD. Error on AGD: < 10%. Before optimization the CNR value was very high for thin PMMA phantoms and hence exposure could be reduced. The AEC of System C2 permitted to set the dose level for every phantom thickness independently. Before optimization the system showed high CNR for PMMA 5cm. Therefore it was possible to reduce exposures in this thickness range. In System C1 a partially fixed exposition setup (31 kv, Rh/Rh) was applied as suggested in literature 9. As can be evinced from Figure 7, for thicknesses 3 cm it was possible to make both CR systems work with a low AGD compared to the other systems, still featuring a satisfying image quality; on the contrary, a higher AGD was required to reach achievable image quality when the phantom thickness increased. System C2 with fixed Mo/Mo target/filter combination requested the highest exposure. Proc. of SPIE Vol J-8

9 Figure 7. Left: Calculated AGD for exposures with automatic exposure control (AEC) at different phantom thicknesses for all systems after optimization. Error on AGD: < 10%. Right: Corresponding CNR rel [%] values Data corresponding to AGD ca mgy System A2 Mo/Mo 30 kv 59 mas System A3 W/Rh 29 kv 129 mas System B1 Rh/Rh 29 kv 70 mas System B2 Rh/Rh 31 kv 60 mas System C1 Rh/Rh 31 kv 56 mas System C2 Mo/Mo 28 kv 70 mas 60 A2 A3 B1 1 B2 C1 C2 Figure 8. Comparison of image quality in different full field mammography systems by CD evaluation. Phantom composition: 2 cm PMM + CDMAM 3.4 detail plate + 2 cm PMMA. Results are given in terms of the overall image quality index calculated by the software CDMAM Analyser, ver. 1.1 (both Artinis, The Netherlands). Data refer to comparable AGD (ca mgy) after system optimization. Systems B1 and B2 permitted only limited optimization by switching manually between the possible AEC modalities: dose (DOSE) for PMMA < 4 cm, standard (STD) for PMMA between 4 and 6 cm. For thicknesses above 6 cm PMMA, which corresponds to a compressed breast thickness of > 7.5 cm (50% glandular, 50% adipose), the AEC was not calibrated. Exposure parameter had to be set by the operator. It was possible to obtain achievable image quality for every phantom thickness, reducing the initially large variations in CNR. This brought dose savings in particular for small phantom thickness, whereas at higher thickness it was necessary to increase exposure to hit image quality requirements. Systems A1 and A2 did not permit any modification in the AEC exposure settings. They used Mo/Mo up to 5 cm PMMA. At large PMMA thickness the CNR values laid beneath the target value. Corresponding dose levels were relatively low, whereas they were relative high at small thickness. Recently mammography units of that model are equipped with a different tube (W/Rh and W/Ag as target/filter combinations), which permits a higher image quality/agd ratio 5. System A3 applied the target/filter combination W/Rh for every phantom thickness. The obtained images always exhibited a CNR well above the target value, requiring at the same time one of the lowest AGD values. Proc. of SPIE Vol J-9

10 4. DISCUSSION 4.1 MTF, NNPS and DQE evaluation Although CR systems turned out to be very low noise systems at higher frequencies, in both studied equipments the intrinsic image quality expressed by DQE reached 40% and confirmed to be substantially lower than in DR units because of limited spatial resolution. Direct conversion detectors demonstrated their superior resolution capabilities compared to indirect conversion technology. Systems A1 and A2 exhibited the highest MTF values over the whole frequency range. That might be correlated to the fact that the corresponding detectors had the smallest pixel size. All 3 systems using a-se (A1, A2, A3) were characterized by high noise levels, as known. The NNPS stayed high at increasing frequency, but the high MTF counterbalanced the poor noise behavior. Resulting maximum DQE laid above 60% in systems A1 and A2. System A3 which features a newer detection system reached the highest peak DQE (> 70%), but its value decreased more rapidly with increasing frequency compared to Systems A1 and A2. Referring to the indirect conversion technology with CsI, NNPS data were available only for System B2, which did not have the latest developments of CSI technology implemented on. The NNPS value diminished with growing frequency, compensating to some extent the lower MTF performance. The calculated DQE reached 50%. 4.2 Image quality assessment Correspondence between the two methods used for image quality assessment, CNR measurement and CD evaluation, has been confirmed. The calculated CNR value and the overall image quality index determined by the commercial software CDMAM Analyser, ver. 1.1 correlated linearly in all studied systems over a very wide range of exposure settings, clinically not used ones included. For equivalent phantom thicknesses > 3 cm PMMA in all systems the target/filter combinations Rh/Rh or W/Rh, which generate x-ray spectra of higher mean energy than Mo/Mo, scored a higher at equal AGD. 4.3 Mammography system optimization The investigated CR systems allowed the widest freedom in exposure parameter setting during optimization, corresponding to the largest relative patient dose reduction, in particular in the small breast sector (PMMA phantom thickness 4 cm). Vice versa, as known from literature, for thickness > 5 cm, it was necessary to increase exposure to come within reach of the CNR target value. Among the DR equipments AEC technique had to be adjusted, too. In all systems except System A3 for small phantom thickness, prior to optimization, the delivered exposure was higher than the necessary level to achieve the desired image quality. AGD savings could be achieved by reducing the load and changing the target/filter choice, but not all systems allowed AEC modifications. System A3, which used W/Rh as target/filter combination for every phantom thickness, performed best in clinical setting with on overall the highest ratio IQF 2 inv /AGD [mgy -1 ]. 5. CONCLUSIONS The higher DQE values featured by all DR units in comparison with the two studied CR systems traduced into significantly higher scores of the global image quality index. It was confirmed that, generally, switching to target/filter combinations that produce an x-ray spectrum of higher mean energy permits dose savings in terms of AGD at equal image quality. However, some systems did not allow any or permitted just limited modifications of the AEC. Hence it was not possible to force all AEC devices to adopt for each phantom thickness the optimal exposure technique. In particular, for low PMMA thickness in all systems but one the applied exposures in AEC modality exceeded the values needed to meet the Euref standards for achievable image quality. In the clinical use System A3, a unit with a-se detector and W target only recently introduced on the market, produced on average the best compromise between image quality and AGD. Its factory AEC settings turned out to be already optimized. REFERENCES 1. Young, K.C., Cook, J.J.H. and Oduko, J.M., Automated and Human Determination of Threshold Contrast for Digital Mammography Systems, IWDM LNCS 4046, (2006) 2. Gennaro, G. and Maggio, C., Dose comparison between screen/film and full-field digital mammography, Eur. Radiol. 16, (2006) Proc. of SPIE Vol J-10

11 3. Monnin, P., Gutierrez, D., Bulling, S., Guntern, D. and Verdun, F.R., A comparison of the performance of digital mammography systems, Med. Phys. 34 (3), (2007) 4. Bernhardt, P., Mertelmeier, T., Hoheisel, M., X-ray spectrum optimization of full-field digital mammography: Simulation and phantom study, Med. Phys. 33 (11), (2006) 5. Baldelli, P., Phelan, N., Egan, G., Effect of Anode/Filter Combination on the Dose and Image Quality of a Digital Mammography System Based on Amorphous Selenium Detector, IWDM LNCS 5116, (2008) 6. Varjonen, M., Strömmer, P., Optimizing the Target/Filter Combination in Digital Mammography in the Sense of Image Quality and Average Glandular Dose, IWDM LNCS 5116, (2008) 7. Oduko, J.M., Young, K., Alsager, A., Gundogdu, O., Effets of Using Tungsten-Anode X-Ray Tubes on Dose and Image Quality in Full-Field Digital Mammography, IWDM LNCS 5116, (2008) 8. Burgess, A.E., The Rose Model, revisted, J. Optical Soc. Am 16, (1999) 9. Young, K.C., Cook, J. and Odoku, J.M, Use of the European Protocol to Optimise a Digital Mammography System, IWDM LNCS 4046, (2006) 10. Pascoal, A., Lawinski, C.P., Honey, I., Evaluation of a software package for automated quality assessment of contrast detail images comparison with subjective visual assessment, Phys. Med. Biol. 50, (2005) 11. Young, K.C., Cook, J., Odoku, J.M. and Bosnans, H., Comparison of Software and Human Observers in Reading Images of the CDMAM Test Object to Assess Digital Mammography Systems, Proc. SPIE Medical Imaging 6142, (2006) 12. Van Engen, R., Young, K., Bonmans, H. and Thijssen, M., European Protocol for the Quality Control of the Physical and Technical Aspects of Mammography Screening. Part B: Digital Mammography [European Guidelines for Quality Assurance in Breast Cancer Screening and Diagnosis], 4th Edition, European Communities, Luxembourg (2006) 13. Oberhofer, N., Parruccini, N. and Moroder, E., Image Quality Assessment and Equipment Optimisation with Automated Phantom Evaluation in Full Field Digital Mammography (FFDM), IWDM LNCS 5116, (2008) 14. Rivetti, S., Lanconelli, N., Campanili, R., Bertolini, M., Borasi, G., Nitrosi, A., Danielli, C., Angelini, L. and Maggi, S., Comparison of different commercial FFDM units by means of physical characterization and contrast-detail analysis, Med. Phys. 33 (11), (2006) 15. Blake, P., Lawinski, C., Mackenzie, A., Honey, I., Cole, H. and Emerton, R., Digital Mammography Systems. Hologic Selenia. A technical report, Kcare Report 05084, (2006) Proc. of SPIE Vol J-11