For the evaluation of the organ doses to the breast, thyroid and ovaries, phantom and patient measurements have been compared with MCS. All measurements have been carried out on a Mammomat Inspiration unit (Siemens Medical Solutions, Erlangen, Germany) using a tungsten anode with rhodium filter. The device was equipped with functions for optimisation of compression (OPCOMP®, Siemens Medical Solutions, Erlangen, Germany) and exposure (OPDOSE®, Siemens Medical Solutions, Erlangen, Germany, Mammomat Inspiration Instruction Manual XPW7-330.620.01.01.01). OPDOSE selects an optimal combination of tube voltage (kV) and anode-filter combination, based on the compression force and compression thickness found by OPCOMP and an automatic exposure control image (short exposition before main image acquisition). With increasing thickness, OPDOSE regulates kV up to reduce time and dose.
Phantom study
The aims of phantom measurements were: (1) to determine the dependence of scatter on compression-thickness, useful to compare surface dose with MCS and patient study; and (2) to evaluate the spatial/angular scatter contributions.
For the first point, an anthropomorphic Alderson phantom [12] (Fig. 1) with a special breast extension for compressed breast has been used (Fig. 1a): slabs of polymethylmethacrylate (PMMA) of 17 × 14 cm in size and thickness of 4, 5, 6 or 7 cm, with the shape of a compressed breast and with breast extension simulating the uncompressed breast. Standard CC and mediolateral oblique (MLO) views were acquired. The dose (air kerma) at the position of the thyroid gland (scatter dose meaning dose related to scattered radiation from breast tissue, collimator, other components and leakage radiation from the unit) has been investigated as a function of the compression thickness by phantom measurements (Alderson phantom with PMMA-slabs for mimicking compressed breasts). The Alderson phantom is designed for radiation therapy and therefore the densities are not exactly tissue equivalent for low x-ray energies [12]. Dedicated MCS of back scatter revealed a correction factor between 1.002 for muscle and 1.03 (28 kVp) or 1.04 (35 kVp) for lead for a scatter angle of 50° (maximum effect at this angle, s. section dosimeter calibration). When the phantom is only used to mimicking the back scatter of a body representing the patient anatomy, this effect can be neglected. The measurements were taken with an RQM solid state detector (IBA Dosimetry GmbH, Schwarzenbruck, Germany) directed to the beam according to Fig. 1 (the RQM sensor does not see the full dose but can indicate the relative increase).
For the second point, to evaluate scatter contributions from breast tissue and collimator system, measurements without Alderson phantom but with PMMA slabs and dosimeters (IBA RQM and re-calibrated Automess SEQ-6R [13], energy range from 18 keV to 3 MeV; section dosimeter calibration) attached to a PMMA plate at thyroid positions have been carried out. To access information about the angular distribution of scattered radiation, measurements have been taken with and without a Pb-shield which covers 180° of the sensitive chamber volume and is directly attached to the dosimeter.
Patient study
The patient study was intended to investigate the influence of variability of patient anatomy on scattered dose in front of and behind the thyroid protection. It was approved by the Ethical Committee at the Kantonsspital Baden, Switzerland, and written informed consent was obtained from all included patients. Patients scheduled for screening or diagnostic mammography were eligible in the absence of the following exclusion criteria: prior operations of one or both breasts; visible asymmetries; palpable lump; and breast implants. Three patients refused participation. One was excluded after inspection (visible breast asymmetry).
Measurements were taken with a modified thyroid collar having a lead equivalent value of 0.25 mm (Wiroma, Niederscherli, Switzerland). A total of 82 patients were categorised in three breast-size categories based on the CC mammogram of the right breast: 27 large (L-group); 22 medium (M-group); and 33 small (S-group). These criteria were based on a volumetric calculation using compression thickness, anteroposterior and right-left dimensions previously measured in 40 mammograms (unpublished data). Eleven patients in the L-group and one patient in the M-group were studied with the with 24 × 30 cm2 paddle; the remaining patients were studied with the 18 × 24 cm2 paddle.
The thyroid collar and two SEQ-6R- dosimeters (see phantom study) were fixed, one in front of and one behind the thyroid collar in the neck midline (Fig. 2). For each patient, the measurements were taken during a two-view standard mammography of the left breast. The application of thyroid collar and dosimeters as well as reading of the results after each measurement were always performed by the same examiner. The images were examined regarding quality and artefacts applying the PGMI-criteria (perfect, good, moderate, inadequate) [14], as required for all certified breast imaging centres in Switzerland. The MGD values were calculated automatically for each exposition by the Mammomat software and were registered for each examination.
To evaluate the protective effect (air kerma behind the collar compared to the air kerma value in front of the collar), a paired t-test was used on the full sample as well as on each of the three breast-size categories (L-group, M-group and S-group). For comparisons among groups, a one-way analysis of variance was used, followed by Tukey’s range test. The statistical tests were applied to the dose values measured in front of the collar as well as to those measured behind the collar. Results with p values < 0.05 were interpreted as statistically significant. Since we expected large breasts to yield more backscatter, we investigated the relationship between body mass index (BMI) (associated with large breasts) and the air kerma in front of the thyroid protection device.
Dosimeter calibration
Because of the high sensitivity at specific low-energy radiation and radial isotropy (ability to measure backscatter), we used Automess SEQ-6R dosimeters (energy range from 18 keV to 3 MeV; dose range 0.01–2.00 mGy). All dosimeters were re-calibrated by separated calibration measurement for the radiation quality in use (30 kVp W-Rh target) to the air kerma with a RQM detector (IBA Dosimetry GmbH, Schwarzenbruck, Germany), with Dosimax plus unit, range 500 nGy – 9999 mGy). Energy dependence and linearity were checked by free air measurements with 20 kVp W-target. We compared the back scatter from muscle tissue according to ICRP110 [15] and muscle tissue under a layer of 0.25 mm Pb (collar lead equivalent) and under a layer of 2.5 mm PMMA, irradiated with 28 and 35 keV W/Rh x-rays by MCS. Based on MCS, the effect of the thyroid collar to the backscatter was estimated to be < 3% and therefore, no separated backscatter correction for the collar was applied for patient measurements. It is assumed that the air kerma (in the investigated energy range, this corresponds to the absorbed dose in air and is in the following taken as measure for the entrance surface dose) at the surface in front of the thyroid was representative for the effect on the thyroid dose.
Monte Carlo simulation
We used the Geant4 simulation toolkit [16]. A numerical voxel phantom was created with the XCAT program [17] (voxel size 2 × 2 × 2 mm3). Each voxel was assigned to an organ. Considering the thyroid gland is as an extended organ with an inhomogeneous dose distribution, the dose was calculated based on the real anatomic situation implemented in the voxel phantom. Accumulated dose was calculated by absorbed energy divided by the voxel mass. The x-ray beam was modelled by electrons with kinetic energies of 28 or 35 keV hitting a W-target inclined at 20°. The radiation was filtered with 50-μm Rh according to the Siemens Mammomat Inspiration manual specifications. The beam opening angle was chosen to fully cover the compression plate. In the model, the x-ray head was simplified as a lead cube of 1-mm wall thickness (shielding of 1E-9 at 35 keV) with a rectangular hole in the bottom. Distances and dimensions were modelled according to the Siemens Mammomat Inspiration manual specifications and on-site measurements. The breast tissue was assumed to be a mixture of fat and glandular tissue (ratio 3:2). According to Verdù et al. [18], different breast tissue compositions have been investigated and the uncertainty of tissue composition onto the glandular dose was estimated to be ± 10% for adipose tissue ratios of 40%, 50% and 60%. From the simulated thyroid dose of approximately 100 pGy, we expected a fluence of 59 cm−2 through the thyroid. Assuming an area of 4 cm2 for the thyroid, we expected 200 photons with an additional uncertainty of 7%. A standard low-energy package was used as recommended in literature [19]. The MCS toolkit was used to calculate all organ doses defined by the ICRP 103 recommendations [20]. The effective dose was calculated on the basis of these organ doses by applying the ICRP 103 model.