1996, The British Journal of Radiology, 69, 555-562. Dose and image quality in mammography with an automatic beam quality system. K C YOUNG, PhD, M L ...
1996, The British Journal of Radiology, 69, 555-562
Dose and image quality in mammography with an automatic beam quality system K C YOUNG, PhD, M L RAMSDALE, MSc and A RUST, MSc National Co-ordinating Centre for the Physics of Mammography, Department of Medical Physics, The Royal Surrey County & St Luke's Hospitals, Guildford GU1 3NT, UK Abstract
Radiation dose, contrast and image quality for automatic beam quality selection (OPDOSE) with the Siemens Mammomat 3000 has been investigated for different breast thicknesses and compared with those found using manually set tube potentials and a molybdenum target and filter. Automatic beam quality selection was found to have a negligible effect for breasts with a compressed breast thickness of less than 45 mm. However, for larger breasts substantial dose savings were achieved for a loss in contrast. For mammograms of compressed breasts with a thickness in excess of 60 mm the mean glandular dose (MGD) per film was 2.90 mGy for manually selected tube potentials with a molybdenum/molybdenum target filter combination as compared with 1.87 mGy using 26 kVp and a tungsten target with rhodium filtration. The contrast loss in using OPDOSE was measured with a test object to be about 10% for breast thicknesses in excess of 45 mm. The standard breast model, which assumes a 50% glandular content, did not provide a good fit to the MGD for women attending for breast screening in the age range 50 to 64 years.
The use of mammography in the screening of well women for breast disease has become very common in many parts of the world. In particular there has been a National Breast Screening Programme (NHSBSP) in the United Kingdom since 1988. Currently more than a million women undergo mammographic examinations each year. For such a large programme, involving a substantial proportion of the healthy population, it is particularly important that the use of radiation be optimized and fully justified. In other words, the need to obtain quality images in order to maximize the sensitivity of cancer detection has to be balanced against the radiation dose to the breast and the consequent risk of cancer induction. One of the main factors affecting image quality and radiation dose is the beam quality used in mammography. In recent years the conventional procedure in the UK has been to use a molybdenum anode X-ray tube, molybdenum filter and a tube potential of 28 kVp (28 kVp Mo/Mo). This was felt to provide X-rays of the appropriate energy and adequate intensity. However, it is well known from theoretical considerations that the optimal energy is different for breasts of different size and composition [1,2]. The three main factors determining the energy distribution of emitted X-rays are the target material, filter material and applied tube potential. Several authors have shown, using mathematical models, that the use of alternative target and filter combinations can achieve significant dose reductions and that 28 kVp Mo/Mo does not provide the optimal beam quality for large or dense breasts and can result in relatively high Received 8 August 1995 and in final form 14 February 1996, accepted 26 February 1996. Vol. 69, No. 822
doses [3, 4 ] . Therefore, a better method of optimizing the radiation dose to this subgroup would be of major interest. The ideal system would be to use a beam quality appropriate for each individual breast. It has not been practical to do this manually with the standard mammographic X-ray set and almost all patients have been exposed with the same factors (28 kVp Mo/Mo). In some centres a higher tube potential has been used for large or dense breasts when long exposure times become a problem. However, X-ray manufacturers are now supplying mammographic X-ray systems with the ability to select a beam quality tailored to each patient with the potential for reducing dose. The objective of this paper is to evaluate the effect of automatic beam quality selection with reference to one such system, the Siemens Mammomat 3000 (M3000). Radiation dose, contrast and overall image quality for automatic beam quality selection with this system have been investigated for different breast thicknesses and compared with those found using conventional technique. Two recognized methods of assessment of patient dose in mammography are used in the UK, and the details of these methods are described in IPSM Report 59/2 [ 5 ] . In the first method, dose is calculated for a standard breast model which comprises a central region with a 50:50 mixture by weight of adipose and glandular tissue and a superficial region of adipose tissue 5 mm thick. The mean glandular dose (MGD) has usually been calculated for a standard breast thickness of 45 mm, but the tables included in the recent revision of IPSM Report 59 [5] allow MGD to be calculated for any thickness. MGD for these standard breasts are estimated from measurements with various thicknesses of Perspex and 555
K C Young, M L Ramsdale and A Rust
by using a table relating Perspex thickness to the equivalent thickness of standard breast. Alternatively, doses on real patients may be estimated by using the post exposure mAs to estimate incident air kerma for each breast and using the same conversion factors in IPSM Report 59/2 [5] to calculate MGD. These conversion factors assume that breast composition is that of the standard breast. In this study both methods have been used to compare the doses for patients radiographed under automatic beam quality selection against doses for patients radiographed using conventional technique (28 kVp Mo/Mo). The major effect on overall image quality expected of changing beam quality is to change the radiological contrast, and this study therefore concentrated on measurements of contrast. This approach has the advantage that contrast measurements are objective and simple to conduct. To assess the link between these contrast measurements and overall image quality, simultaneous measurements were made of contrast and image quality, using an image quality test object and Perspex blocks to simulate small, medium and large breasts. Materials and methods Equipment The Siemens M3000 has an X-ray tube with molybdenum and tungsten targets and molybdenum and rhodium niters. Automatic selection of the appropriate target, filter and tube potential is based on the compressed breast thickness derived from the position of the compression plate prior to exposure. The four tube potential target/filter combinations selected in automatic mode (OPDOSE) for different compressed breast thickness for the system tested are shown in Table I. For all procedures Fuji UM-MA film was used with Fuji UM-MA fine screens in Fuji EC-MA carbon fibre cassettes. Films were processed using Photosol chemicals in a Fuji FPM 3000 processor with a developer temperature of 33.5° and an extended cycle of 210 s dry-to-dry. Dose to the standard breast The method described in IPSM Report 59/2 [5] for the assessment of dose to the standard breast was followed. The incident air kerma (K) required to achieve an optical density of 1.60 (including base plus fog) was determined for Perspex blocks of thickness ranging from 20 to 70 mm. The blocks used had been machined to an accuracy of ±0.1 mm. The half value layer (HVL) for each beam quality was measured as described in the
same report [ 5 ] . The MGD for the thicknesses of the standard breast model equivalent to each thickness of Perspex tested was calculated from K and HVL using the tables provided in IPSM Report 59/2 [5]. Dose measurement from patient data Doses per mammogram were estimated for the Siemens M3000 at an assessment clinic in a screening centre. Post-exposure mAs, breast thickness, tube potential, target and filter were recorded for 169 mammograms using OPDOSE. Incident air kerma were estimated from the post exposure mAs and measurements of output per mAs for each target/filter combination and corrected for individual breast thicknesses and attenuation by the compression plate. MGD for each breast was calculated from the incident air kerma using the conversion factors in the table in IPSM Report 59/2 [5] which are based on the simplifying assumption that the breasts have the same composition as the standard breast. During the period of dose assessment the target optical density was 1.6 + 0.15. However, unlike the measurements with Perspex blocks, doses for individual mammograms were not corrected to a standard optical density and were subject to variations in automatic exposure control (AEC) performance. The breast thickness used was that recorded on the digital display of the M3000, which may be subject to some error due to the variable flexing of the compression plate from one examination to another. To assess the patient doses using a manually set tube potential it would have been ideal to use the same system. However, this was not possible for logistical reasons and doses were instead determined using a Siemens Mammomat 300 (M300) mammography X-ray set. The M300 was on a mobile unit of the same screening centre, and data on 253 mammograms for women in the age range 50-64 years were obtained using a manually set tube potential with a molybdenum target and filter. The M300 is virtually identical to the M3000 except for the lack of the OPDOSE feature. The method of dose estimation was as described for the M3000. However, it cannot be assumed that the doses for these two systems would be identical even if the same technique were used. The measurements of tube potential accuracy, grid factor, filtration and breast thickness and target optical density were all identical within experimental error but the output per mAs was about 20% higher for the M3000. Estimates of the mean MGD to the standard breast derived from measurements with Perspex for the two systems showed some variation during the clinical
Table I. Beam quality factors selected by Siemens Mammomat 3000 in automatic mode Compressed breast thickness
kVp
Target material
Filter material
0-30 mm 30-45 mm 45-60 mm >60 mm
26 27 27 26
Molybdenum (Mo) Molybdenum (Mo) Molybdenum (Mo) Tungsten (W)
30 fim 30 um 25 um 50 um
556
Measured HVL with compression plate (mm Al)
molybdenum (Mo) molybdenum (Mo) rhodium (Rh) rhodium (Rh)
0.33 0.34 . 0.40 0.51
The British Journal of Radiology, June 1996
Dose and image quality in mammography
evaluation. The MGD from the M300 was generally about 10% higher, probably due to a combination of reciprocity effects andxassette speed differences. To compare the doses estimated from patient data with those from Perspex block data the mAs selected by the AEC of the M300 were recorded daily for 20 mm, 40 mm and 60 mm thick Perspex blocks for 28 kVp set. These AEC selected mAs values were used to calculate the MGD for the equivalent thicknesses of standard breast without any film density correction. These MGD estimates for the standard breast composition were then compared with the estimates from patient data from the M300. Theoretical estimation of contrast Emission X-ray spectra for the four anode target/filter combinations provided under automatic beam quality selection on the M3000 (Figure 1) were generated using a program derived from Birch and Marshall [ 6 ] . The contrast in energy fluence for a 0.2 mm thickness of aluminium placed on 20-70 mm thicknesses of Perspex was calculated from the photon spectra for each target filter combination and the linear attenuation coefficients of Perspex and aluminium at 0.5 keV intervals. A method similar to that described previously by Desponds et al [7] was used. The effect of scatter was included by using contrast degradation factors estimated from the scatterto-primary ratios for a moving grid calculated by Dance et al [8]. Film contrast was derived from the contrast in energy fluence and the film gradient for the filmscreen system in use. Contrast measurement To measure contrast a test object comprising a 10 mm square sheet of 0.2 mm thick aluminium, placed on top of a 20 mm stack of Perspex, was used in a manner similar to that described previously [ 7 ] . The aluminium square was positioned on the midline and centred at 60 mm from the chest wall edge of the film to avoid the AEC detector. The stack was radiographed using both conventional technique and automatic beam quality selection. Additional 10 mm thick sheets of Perspex were added to simulate increasing breast thickness and further
26 kVp Mo/Mo 27 kVp Mo/Mo 27kVpMo/Rh 26 kVp W/Rh energy fluence
0
5
10
15
20
photon energy (keV)
Figure 1. Normalized emission spectra for the four target filter combinations used by the OPDOSE system on the Siemens Mammomat 3000.
Vol. 69, No. 822
radiographs obtained up to a total of 70 mm Perspex. Contrast was measured as the difference in optical density between the image of the aluminium square and the adjacent background. Several density measurements were averaged to determine these densities. Errors for contrast are quoted as ±0.01 OD and were largely determined by the precision of the densitometer. The variation of contrast with optical density was also measured and used to normalize the measurements of contrast to that expected where the background density was 1.60 including base plus fog. This was the target optical density in use at the screening centre, and corresponds to the midpoint of the recommended optical density range from 1.40 to 1.80 in the UK NHSBSP [ 9 ] . Effect of tube potential changes The effect of tube potential changes for a large breast were simulated using 70 mm of Perspex. MGD and contrast were measured for three different tube potential settings for each of the three target and filter combinations. Overall image quality measurement Overall image quality was assessed using a TOR(MAM) mammography test object, the use of which has been described elsewhere [10, 11]. Image quality was measured for the conventional and automatic beam quality selections at three different simulated thicknesses of breast: (a) small: TOR(MAM) on top of 10 mm Perspex; (b) medium: TOR(MAM) on top of 30 mm Perspex; (c) large: TOR(MAM) with 30 mm Perspex below and 25 mm of Perspex on top. The test object itself is equivalent to about 15 mm of Perspex. A 0.2 mm thick aluminium square was taped on the midline of the TOR(MAM) test object, avoiding critical test details within the object, to allow contrast to be measured on each image quality film. Image quality scores were averaged for three observers and the errors calculated as + 2 standard errors of the mean. Apart from the exceptions noted above care was taken to ensure that measurements of MGD, contrast and image quality were normalized to a fixed optical density of 1.60. Results Dose to standard breast MGDs for various thicknesses of the standard breast composition, estimated from measurements with Perspex blocks, are shown in Figure 2 for 28 kVp Mo/Mo and OPDOSE settings of the M3000. Corresponding exposure times for each thickness of the standard breast tested for the two modes are shown in Table II. Up to a thickness of 45 mm, both modes select molybdenum as the target and filter material, and the slightly lower tube potential selection in the OPDOSE mode resulted in only a small increase in dose. However, for greater thicknesses the OPDOSE mode selected different target and filter combinations resulting in significantly reduced doses with savings of 27%, 34% and 59% for thicknesses of 54 mm, 64 mm and 75 mm, respectively. The backup timer terminated exposures at 4.0 s limiting the
557
K C Young, M L Ramsdale and A Rust Table II. Exposure times for Siemens Mammomat 3000 for a gross film density of 1.60 Perspex thickness (mm)
Equivalent thickness of standard breast (mm)
Exposure time using OPDOSE (s)
Exposure time using 28 kVp Mo/Mo (s)
20 30 40 50 60 70
23 33 44 54 65 75
0.09 0.21 0.42 0.64 1.14 3.03
0.07 0.16 0.35 0.73 1.69 3.45
8 7 6-
— • — 2 8 kVp Mo/Mo - - • - OPDOSE
5 MGD
4
(mGy)
3 27 kV Mo/Mo
2 1
26 kV Ma/Mo
26 kV Ma/Mo.
A"
27kV Ma/Rh
26 kV W/Rh
27 kV Mo/Rh
20 40 60 thickness of standard breast (mm)
80
Figure 2. Mean glandular doses versus thickness of standard breast for the Siemens Mammomat 3000 using 28 kVp Mo/Mo and OPDOSE.
maximum dose to 8 mGy using 28 kVp Mo/Mo, and to 4 mGy using OPDOSE. Patient doses MGDs per film estimated from patient data using conventional technique on the M300 are shown in Figure 3. The majority of women (236 mammograms) were screened using 28 kVp but 30 kVp was chosen for a few with particularly large breasts (16 mammograms).
Mean MGD was 2.15 mGy for all mammograms and 2.90 + 0.08 (s.e.m.) mGy for 123 mammograms of breasts with a compressed thickness of greater than 60 mm, with a mean thickness of 69.7 + 0.7 mm. For the M300 system the MGD for a standard breast (45 mm thickness) was estimated to be 1.44 mGy on the 2 days of the patient dose measurements. MGDs per exposure estimated for 169 mammograms using OPDOSE on the M3000 are shown in Figure 4. Mean dose was 1.38 mGy for all mammograms and 1.87 + 0.10 (s.e.m.) mGy for 67 mammograms of breasts with a compressed thickness greater than 60 mm, with a mean thickness of 66.5 + 0.8 mm. The doses for the standard breast model are also shown as a line superimposed on this graph. MGD was estimated to be 1.30 mGy for a 45 mm thick standard breast. Patient doses for automatic beam quality selection using the M3000 are compared with those for conventional technique using the M300 in Figure 5, by integrating the data into 5 mm bands of compressed breast thickness. Mean doses corresponding to the last three data points for the M3000 are based on only one, two and two mammograms, respectively, and are therefore subject to larger errors than the other data points. Comparison of dose measurement methods A curve representing the MGD calculated for the standard breast model is shown for comparison with the
27 kV Mo/Rh 26 kV W/Rh 26 kV Mo/Mo 27 kV Mo/Mo -standard breast
7 6
mgd (mGy)
5 MGD (mGy)
4
3 2 1 -
0 40
breast thickness (mm)
Figure 3. Mean glandular doses per film from patient data for the Siemens Mammomat 300 using Mo/Mo target and filter. The doses predicted from Perspex data for the standard breast using 28 kVp Mo/Mo are shown as a dashed curve.
558
20
60
Figure 4. Mean glandular the Siemens Mammomat tube potential and target doses predicted for the Perspex data is overlaid.
40 60 80 compressed breast thickness (mm)
100
doses per film from patient data for 3000 using OPDOSE to select the filter combination. A line showing standard breast composition from
The British Journal of Radiology, June 1996
Dose and image quality in mammography 6 0.35 -
5 MGD (mGy)
4
contrast (OD) « 28 kVp Mo/Mo A 30 kVp Mo/Mo • o-OPDOSE
0.30 0.25
measured 28 kV Mo/Ma •theoretical 28 kV Mo/Mo measured OPDOSE
0.20 --
3
-theoretical OPDOSE
0.15 --
H
26 kV W/Rh
26 kV Mo/Mo
0.10
OPDOSE Selections 0.05 0.00 20
breast thickness (mm)
Figure 5. Average MGD per film for patient data grouped by compressed breast thickness in 5 mm bands for conventional technique using the Siemens M300 and using OPDOSE on the Siemens M3000. Table III. Estimates of equivalent breast thickness of Perspex from various studies Thickness of Perspex (mm)
20 30 40 50 60 70
Equivalent breast thickness (mm) Standard breast [5]
Breasts in this study
Reference [12]
22.7 33.2 43.6 54.1 64.6 75.0
20 34 47 61 75 89
10 30 50 70 90 110
patient data in Figures 3 and 4. In addition, the mAs values selected by the AEC for the Perspex blocks were compared with the distribution of post-exposure mAs values recorded for different thicknesses of compressed breast from the patient series. A Perspex-to-breast equivalence table for the M300 patient data was derived, as shown in Table III alongside the values reported in other studies. Contrast Theoretical and measured contrasts for the Mo/Mo combination at 28 kVp and the alternative automatic beam quality selections under OPDOSE mode for different thicknesses of Perspex are shown in Figure 6. There was good agreement between theory and experimental results and for 28 kVp Mo/Mo the contrast fell from a maximum of 0.38 to 0.25 as Perspex thickness increased from 20 to 70 mm. In the OPDOSE mode the contrast fell from 0.38 to 0.23 over the same range of breast thickness. The effect on contrast of using OPDOSE instead of 28 kV Mo/Mo was calculated from the theoretical and experimental data shown in Figure 6. Errors in the experimental estimates of these contrast differences were estimated from the uncertainty in the contrast measurements of +0.01. Theory predicted that for thicknesses of 20, 30 and 40 mm of Perspex the
Vol. 69, No. 822
30 40 50 Perspex thickness (mm)
Figure 6. Contrast versus Perspex thickness using 28 kVp Mo/Mo and the OPDOSE facility on the Siemens Mammomat 3000.
OPDOSE mode would yield a relative increase in the contrast of 0%, 1% and 2%, respectively. Corresponding increases in contrast of 0 + 4%, 3 ± 4 % and 7 + 5% were recorded experimentally. For thicknesses of 50, 60 and 70 mm the model predicted that the OPDOSE system would result in contrast reductions of 10%, 12% and 8%. These closely matched the measured contrast reductions of 9 + 5%, 10 + 5% and 10±6%. MGD and contrast for a large standard breast, equivalent to 70 mm Perspex, for various combinations of target, filter and tube potential on the M3000 are shown in Figure 7. The OPDOSE mode selected 26 kVp W/Rh for 70 mm Perspex, and the other combinations were set manually to assess their relative effect on dose and contrast. The highest dose of 6.8 mGy was measured using 28 kVp Mo/Mo, and this beam quality also resulted in the highest contrast of 0.24 + 0.01. Using 30 or 32 kVp resulted in lower doses of 5.0 mGy and 3.6 mGy but with lower contrast of 0.22±0.01 and 0.20 + 0.01, respectively. Using 27 kVp Mo/Rh resulted in a dose of 4.9 mGy with essentially the same contrast as 28 kVp Mo/Mo. Much the lowest dose was achieved using the 7 r 6 MGD (mGy) 5
4 3
W/Rh *
2
30 kV W/Rh
- * 28 kV W/Rh
1 error in contrast
0.17
0.19
0.21 contrast
0.23
0.25
Figure 7. Dose versus contrast for 70 mm Perspex using Mo/Mo, Mo/Rh and W/Rh target filter combinations on the Siemens Mammomat 3000.
559
K C Young, M L Ramsdale and A Rust
breast model showed a dose saving which increased from approximately 30% at a thickness of 50 mm to about 60% at 75 mm thickness. As regards the patient data 28 kV Ma/Mo 28 kV Mo/Mo 80 some caution must be exercised in comparing the doses 70 mm thick from the M300 and M3000. Although these are very TOR(MAM)) 26 kV W/Rh score 70 mm thick similar machines the MGDs for standard breasts were 27 kV Mo/Rh 60 not identical, even for the same technique. Thus, differences in the doses for the two systems will be partly due 40 to different beam quality selection and partly due to other factors such as cassette speed, AEC drift and 20 differing reciprocity effects. However, the general pattern found for the standard breast was confirmed from the patient data with the little difference in dose for smaller 1 1 1— 1 0 breasts but with an increasing dose saving the larger the 0.00 0.10 0.20 0.30 0.40 breast. For women with compressed breast thickness in contrast (OD) excess of 60 mm the average MGD was 2.90 mGy using Figure 8. Contrast and image quality score for different target the conventional technique on the M300 (28 kVp filter combinations and Perspex thicknesses for the Siemens Mo/Mo) as compared with 1.87 mGy using OPDOSE Mammomat 3000. on the M3000. The use of automatic beam quality selection also Table IV. Effect of target and filter selection on dose and image affected the contrast. Experimental results showed good quality for 70 mm Perspex using Siemens Mammomat 3000 agreement with the theoretical predictions, confirming expectations of the reduction in contrast due to increased Dose Contrast Image quality Factors beam hardening and scatter with increasing thickness or (mGy) (OD) TOR(MAM) score density of breast. For breast thicknesses less than 45 mm 28 kVp Mo/Mo 6.8 0.21 ±0.01 71 ± 2 the effect of using OPDOSE on contrast was minimal, 27 kVp Mo/Rh 4.9 0.22 + 0.01 69 + 2 as it was for dose. However, for thicknesses greater than 26 kVp W/Rh 2.8 0.18 + 0.01 67 + 2 45 mm there was a loss in contrast of about 10%. Thus the reduction in dose demonstrated for large breasts using OPDOSE was accompanied in this study by a W/Rh target/filter combination and using 26 kVp W/Rh relatively small reduction in radiographic contrast. resulted in a dose of 2.8 mGy with a contrast of The TOR(MAM) image quality test object was used 0.22 + 0.01. The use of higher tube potential (28 and as a measure of overall image quality to investigate the 30 kVp) with the W/Rh target filter combination had significance of contrast loss. A linear relationship was little effect on dose and the small decreases in contrast found between the TOR(MAM) score and contrast appeared insignificant and within experimental error. (Figure 8). This shows the validity of using a simple The correlation between measured contrast and TOR(MAM) scores is illustrated for various thicknesses contrast measure as a surrogate measure of overall image of Perspex and target filter combinations in Figure 8. quality in these experiments. In addition to the expected Differences in thickness had a marked effect on contrast reduction in image quality score and contrast with and overall image quality, with the smallest Perspex increasing thicknesses of breast, a loss in contrast of stack having the best image quality. For the simulated about 12% was measured when 26 kVp W/Rh was used small breast (25 mm Perspex) there was no significant instead of 28 kVp Mo/Mo for a large breast simulated difference in image quality between the OPDOSE and by 70 mm Perspex. This loss in contrast corresponded conventional modes, with image quality scores of 92 + 10 to a reduction in image quality score from 71 + 2 to and 90 + 7, respectively. For simulated large breasts there 67 + 2, accompanied by a dose reduction of 59% was no significant difference in overall image quality (Table IV). The use of 27 kVp Mo/Rh for the simulated between 28 kVp Mo/Mo and 27 kVp Mo/Rh with scores large breast involved no significant reduction in contrast of 71 + 2 and 69 + 2, respectively. However, using 26 kVp or image quality but achieved a 29% reduction in dose. W/Rh did result in a lower image quality score of 67 + 2. The authors have previously reported TOR(MAM) Table IV summarises the corresponding dose, contrast scores that ranged from 54 to 88 for 41 mammography and image quality score for a large breast, simulated by systems at NHS breast screening centres [13]. Thus the 70 mm Perspex, for three different beam quality selec- reduction in image quality for large breasts using OPDOSE can be expected to be much smaller than the tions using the M3000. typical variation from system to system. Any significant reduction in the film density would also have a much Discussion The application of automatic beam quality selection larger effect on contrast and image quality than these has been shown in this study for one model of X-ray set changes in beam quality [ 9 ] . Nonetheless, any deteriorto reduce the dose to the patient in mammography. ation in image quality is undesirable and deterioration Estimates of MGD using Perspex data and the standard from this source may still be clinically relevant. 100
560
28 kV Mo/Mo 45 mm thick T
26 kV Mo/Mo 25 mm thick
The British Journal of Radiology, June 1996
Dose and image quality in mammography
The conventional method of increasing X-ray penetration in large or dense breasts using a Mo/Mo target/ filter combination has-been to increase the tube potential. Some X-ray manufacturers have designed automatic systems which rely on tube potential changes alone to select beam quality. The question therefore arises whether this method of adjusting the beam quality is adequate and whether changes in target and filter materials yield any additional benefits. The effect of changing tube potential alone for the three target filter combinations available was investigated for a large breast simulated by a 70 mm thickness of Perspex. As can be seen from Figure 7 increasing tube potential with a Mo/Mo target and filter combination did reduce dose, but resulted in a greater contrast loss for the same dose saving achieved by changing to the other target filter combinations. For example, using 30 kVp Mo/Mo required a dose of 4.9 mGy, as compared with using 26 kVp W/Rh which required a dose of 2.8 mGy for the same contrast. It should be borne in mind that the method of beam quality selection used by the M3000 is based on compressed breast thickness alone. Such a system cannot take account of the large variations in breast density due to the varying glandular content of breasts and the beam quality selection may not always be fully optimized. Some other manufacturers base automatic beam quality selection on the transmitted radiation detected by the AEC sensor in either a pre-exposure pulse or at the beginning of the main exposure. Such designs may enable more fully optimized choices to be made for individual breasts. The MGD calculated for the standard breast model did not provide a good fit to the patient data in either automatic or manual modes. This has also been reported in another study [12]. The patient data used here were for women attending for screening in the age range 50-64 years and it can be expected that for these postmenopausal women the average glandular content was much less than the 50% of the breast tissue assumed in the standard breast model. A comparison of the mAs selected by the M300 AEC for different thicknesses of Perspex and compressed breast allowed a table of conversion from Perspex thickness to breast equivalent thickness to be calculated for this patient group (Table III). This resulted in equivalent breast thicknesses which were greater than those estimated for breasts with a 50% glandular content in IPSM Report 59/2 [5] but less than those reported using patient data by Thilander et al [12]. The effect of this difference between the standard breast model and breasts in the screened population is that the standard breast model leads one to expect higher incident air kerma and therefore greater MGD than is found in practice with the much more fatty breast screened in older women. This effect is quite small for the 45 mm thick standard breast model but for large breasts the incident air kerma may be overestimated by a factor of about 2. Regardless of how the incident air kerma was estimated, all MGDs calculated in this study used the conversion factors quoted in IPSM Report 59/2 [5] for the standard breast with a 50% glandular content. As the
Vol. 69, No. 822
glandular content was probably much less than this for women attending for breast screening, different conversion factors would be more appropriate for these women. It can be seen from recently published tables of conversion factors that assuming a lower glandular content of say 25% would increase the conversion factors for large breasts by about 20% [14]. Conclusions
For larger breasts substantial dose savings were made using the alternative anode and filter combinations of the automatic beam quality mode on the Siemens M3000. However, for small to average breasts the effect of using the automatic beam quality facility as compared with the conventional technique was negligible. The dose reductions with large breasts were achieved at a small cost in terms of reduced contrast. Gains in contrast achieved by using tube potential settings of less than 28 kVp for the thinner breasts were small. For all breasts the highest contrast and therefore the best overall image quality was achieved by using a molybdenum target and filter combination. Increasing the applied tube potential for large breasts without changing the filter or target was found to be a less than optimum method of dose reduction since it caused a relatively greater loss in contrast than changing filter/target. The clinical significance of the small contrast losses that accompany the dose savings from automatic beam quality selection remains debatable and whether they are acceptable may depend on the application. The standard breast model, which assumes a 50% glandular content, did not provide a good fit to the MGD for women attending for breast screening in the age range 50-64 years. This was particularly noticeable for larger breasts. There appears from this study to be a strong case for further research into the physical characteristics of breasts examined in breast screening, with the aim of deriving a more appropriate breast model for dosimetry in the UK NHSBSP. Acknowledgments
We are grateful to the staff of the Jarvis Breast Screening Centre in Guildford for their assistance in conducting this work. In particular we would like to thank Bobby Thomas and Amanda Chase for collecting the data used to calculate patient doses. References 1. JENNINGS, R J and FEWELL, T R, Filters—photon energy control and patient exposure. In Reduced Dose Mammography, ed. by W W Logan and E P Muntz) (Masson Publishing, New York), pp. 212-222 (1979). 2. DANCE, D R and DAY, G, Simulation of mammography by Monte Carlo simulation—the dependence of radiation dose, scatter and noise on photon energy. In Patient Exposure to Radiation in Medical X-ray Diagnosis, ed. by G Drexler, H Eriskat and H Schibilla (CEC, Brussels), pp 227-243. (1981). 3. DESPONDS, L, DEPEURSINGE, C, GRECESCU, M ET AL, Influence of anode and filter material on image quality and glandular dose for screen-film mammography, Phys. Med. Biol, 36, 1165-1182 (1991).
561
K C Young, M L Ramsdale and A Rust 4. STANTON, M S, DAY, J L, VILLAFANA, T ET AL, 10. COWAN, A R, BRETTLE, D S, COLEMAN, N J and PARKIN, G J S, A preliminary investigation of the imaging Screen-film mammographic technique for breast cancer performance of photostimulable phosphor computed radiscreening, Radiology, 163, 471-479 (1987). ography using a new design of mammographic quality con5. INSTITUTE OF PHYSICAL SCIENCES IN trol test object, Br. J. Radiol., 65, 528-535 (1992). MEDICINE, The Commissioning and Routine Testing of Mammographic X-ray Systems, Report 59/2 (IPSM, York) 11. YOUNG, K C and RAMSDALE, M L, Evaluation of mam(1994). mography image quality phantoms, Radiat. Prot. Dosim., 49, 171-173(1993). 6. BIRCH, R, and MARSHALL, M, Computation of Brehmsstrahlung X-ray spectra and comparison with spec12. THILANDER, A, EKLUND, S, LEITZ, W and tra measured with a Ge(Li) detector, Phys. Med. BioL, 24, MATTSON, S, Special problems of patient dosimetry in 505-517(1979). mammography, Radiat. Prot. Dosim., 43, 217-220 (1992). 13. YOUNG, K C, RUST, A and RAMSDALE, M L, 7. DESPONDS, L, KLAUSZ, R and HEIDSIECK, R, Image Mammographic Dose and Image Quality in the UK Breast quality and glandular dose for rhodium K-edge filter in Screening Programme, NHSBSP Publication No 35 mammography: performance comparison of the rhodium (NHSBSP publications, Sheffield) (1995). anode and molybdenum anode, Coolidge Sci. Review, 1, 7-12(1994). 14. WU, X, BARNES, G T and TUCKER, D M, Spectral 8. DANCE, D R, PERSLIDEN, J and ALM CARLSSON, G, dependence of glandular tissue dose in screen-film mamCalculation of dose and contrast for two mammographic mography, Radiology, 179, 143-148 (1991). grids, Phys. Med. BioL, 37, 235-248 (1992). 9. YOUNG, K C, WALLIS, M G and RAMSDALE, M L, Mammographic film density and detection of small breast cancers, Clin. Radiol, 49, 461-465 (1994).
562
The British Journal of Radiology, June 1996
