If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Modeling a breast compressed in MLO view instead of in CC view does not produce any significant differences in MGD estimates.
•
Skin model presents a noticeable influence on T-factor in DBT.
•
Correction factor for taking into account the ion-chamber model in MGD estimates in BCT was calculated.
Abstract
Purpose
To provide mean glandular dose (MGD) estimates via Monte Carlo (MC) simulations as a function of the breast models and scan parameters in mammography, digital breast tomosynthesis (DBT) and dedicated breast CT (BCT).
Methods
The MC code was based on GEANT4 toolkit. The simulated compressed breast was either a cylinder with a semi-circular section or ad hoc shaped for oblique view (MLO). In DBT we studied the influence of breast models and exam parameters on the T-factors (i.e. the conversion factor for the calculation of the MGD in DBT from that for a 0-degree projection), and in BCT we investigated the influence on the MGD estimates of the ion chamber volume used for the air kerma measurements.
Results
In mammography, a model representative of a breast undergoing an MLO view exam did not produce substantial differences (0.4%) in MGD estimates, when compared to a conventional cranio-caudal (CC) view breast model. The beam half value layer did not present a significant influence on T-factors in DBT (<0.8%), while the skin model presented significant influence on MGD estimates (up to 3.3% at 30 degrees scan angle), increasing for larger scan angles. We derived a correction factor for taking into account the different ion chamber volume used in MGD estimates in BCT.
Conclusions
A series of MC code modules for MGD estimates in 2D and 3D breast imaging have been developed in order to take into account the most recent advances in breast models.
]. The MGD in a breast exam can be calculated from measurements of air kerma (mGy) in a fixed area of the field of view and multiplication by suitable conversion coefficients (mGy/mGy): the normalized glandular dose coefficients (DgN). These coefficients are computed via Monte Carlo (MC) simulations and depend on the breast anatomy, scanner geometry and X-ray spectra characteristics. In MC simulations the scanner geometry is replicated and the shape and composition of the breast are suitably modelled. A common model for breast dosimetry includes an internal breast volume composed of a homogeneous mixture of glandular and adipose tissue, surrounded by an envelope of skin tissue. New 3D breast imaging techniques (dedicated breast computed tomography, BCT) [
] provided a detailed 3D description of the breast anatomy in patients, so permitting more realistic breast models for dosimetry in 2D and 3D X-ray breast imaging [
Institute of Physics and Engineering in Medicine (IPEM). The commissioning and routine testing of mammographic X-ray systems. IPEM Report 89 (York, United Kingdom: IPEM); 2005.
]. Such a difference in the skin model (4–5 mm vs. 1.45 mm) may affect the dose estimates derived from DgN calculations. In 2D mammography, the MGD estimate is higher for a thinner skin layer [
]. Moreover, in BCT – where higher photon energies than in mammography are adopted – the skin model presents some impact on the (photon energy dependent) MGD estimates adopted for clinical scans [
Institute of Physics and Engineering in Medicine (IPEM). The commissioning and routine testing of mammographic X-ray systems. IPEM Report 89 (York, United Kingdom: IPEM); 2005.
] breast models for MGD estimates do not consider the heterogeneous distribution of the breast glandular tissue within the organ volume: indeed, they assume that the inner portion of the breast is made of a homogeneous mixture of glandular and adipose tissue. This model simplification may produce a bias in MGD estimates, when compared to estimates based on patient specific structured models derived from clinical BCT scans [
]. In particular, this assumption neglects the large variability in the glandular tissue texture and it is representative of a hypothetical “average” breast [
Another point of interest in DgN calculations for breast dosimetry comes from the observation that the breast models adopted in mammography and DBT are devised for the craniocaudal (CC) view; MGD estimates for mediolateral oblique (MLO) views, for diagnostic views and for particular cases – such as contrast enhanced mammography [
Estimation of mean glandular dose for contrast enhanced digital mammography: factors for use with the UK, European and IAEA breast dosimetry protocols.
In this scenario, the paradigm for X-ray breast imaging dosimetry is under review. The joint AAPM-EFOMP task group 282 “Development of a new – universal breast dosimetry method” is developing “a new model and corresponding methodology to estimate the breast average glandular dose (AGD) from x-ray based image acquisitions including standard mammography, contrast-enhanced mammography, spot mammography, magnification mammography, and breast tomosynthesis” [
]. It is expected that the investigation in this work may represent a contribution to this task.
In previous papers, we adopted various skin models and investigated the influence of physical processes (bremsstrahlung) and electron cut-off energies adopted in MC simulations for DgN calculation in mammography and BCT [
]. In addition, we compared the MDG estimated via homogeneous breast models in mammography and BCT to that obtained with patient specific digital breast phantoms, which reflect the real breast anatomy and shape [
Taking into account all the previous considerations, we upgraded our MC code for MGD and DgN estimates in 2D and 3D X-ray breast imaging: in this work we present the most recent MC code version by our group which includes the possibility of simulating the geometry adopted in DBT scanners. As a specific additional development included in this code, we calculated the MGD and the conversion coefficients in DBT and introduced a breast model representative of the compressed breast in MLO view. Dose estimates based on the adoption of a MLO breast model in 2D mammography have been produced, as well as simulations showing the influence of the beam quality, skin model and geometrical parameter in DBT, and the influence of the ion chamber volume used in measurements for MGD estimates in BCT.
2. Materials and methods
2.1 Monte Carlo code
Our Monte Carlo code was based on the GEANT4 toolkit version 10.00 and relied on the low energy physics list Option4 [
]. Photoelectric, Compton scatter and Rayleigh photon interactions were simulated (without the inter-atomic interference effects in small angle coherent X-ray scattering [
The inner part of the irradiated breast was simulated as a homogeneous mixture of adipose and glandular tissue and the MGD was calculated as reported in Ref. [
where is the energy released at the interaction event i, fg is the breast glandular fraction by mass, and Wb is the breast mass (skin excluded). The factor G(E) [
Here, μen/ρ is the mass energy absorption coefficient of glandular (subscript g) and adipose (subscript a) tissues, evaluated following the functional interpolation in Ref. [
]. The G-factor was evaluated at the energy of the interacting photon. Finally, DgN can be calculated as follows:
(3)
K is the air kerma calculated in a fixed region in the field of view, obtained by scoring the photon fluence passing through a defined part of the irradiated area [
Here, Ei is the energy of the i-th photon which passes through a scoring surface of area S, μen/ρair is the mass energy absorption coefficient of the (dry) air as contained in the NIST database, θi is the angle between the photon direction and the direction perpendicular to the scoring surface. In 2D mammography, the scoring region is usually located at the entrance breast surface [
Institute of Physics and Engineering in Medicine (IPEM). The commissioning and routine testing of mammographic X-ray systems. IPEM Report 89 (York, United Kingdom: IPEM); 2005.
In DBT, where the X-ray source rotates over an arc around the breast, the MGD for a single projection, acquired at a scan angle α (MGDα), can be related to the MGD for a 0 degrees projection by means of the t-factor (relative glandular dose, RGD [
MGD and MGDα are produced with the same tube load (i.e. number of launched photons in MC simulations). α is the angle between the central X-ray beam and the detector plane [
Here, IDD is the isocenter-to-detector distance and SID is the source-to-isocenter distance. Fig. 1a reports a scheme of the irradiation geometry adopted in DBT exams.
Fig. 1Scheme of the irradiation geometry in a) a DBT exam and in b) a BCT exam.
The MGD for the DBT exams (MGDDBT) can be computed by means of the T-factor:
(7)
In Eq. (7), the DgN coefficient is the same as the one adopted for the given breast size and composition and X-ray spectrum in mammography; the air kerma K is measured at 0 degrees projection angle for the total tube load adopted for the considered exam. Then, T-factors [
In BCT, where the uncompressed breast hangs at the scanner isocenter and the gantry rotates over 360 degrees around a vertical axis of rotation, the MGD (MGDBCT) can be calculated from the air kerma measured at the scanner isocenter (Kiso) for a full rotation. Hence, MGDBCT estimates are computed by means of specific DgN coefficients, DgNCT [
], which take into account the scanner geometry, the X-ray spectrum and uncompressed breast shape and composition:
(9)
In BCT, the breast is usually simulated as a homogeneous cylinder or a semi-hellipsoid, and its cylindrical symmetry around the axis of rotation permits to simulate the exam with the X-ray source in a fixed position. KISO is computed in the MC simulations as indicated in Eq. (4) from the scored photon fluence passing through a region at the scanner isocenter (Fig. 1b). In the KISO simulations, the patient breast and body are removed from the geometry.
We assessed that the values simulated in this work presented a statistical uncertainty not exceeding 0.2% of the mean value. It was reached with 108 launched photons and a computational time of less than 1 h at each photon energy.
Data in this work are shown in compliance with the recommendations of the AAPM Task group 268 [
The breast mass was simulated as a homogenous mixture of adipose and glandular tissue enveloped in a layer simulating the skin. The skin was modelled either as 5.00 mm of adipose layer (as proposed in Refs. [
], the X-ray point source was isotropic and electronically collimated in order to irradiate exclusively a detector surface of 24 × 30 cm2 placed at 66.0 cm from the focal spot of the X-ray tube. The top compression paddle was made of 2.8-mm thick polycarbonate and the support paddle was made of 4.1-mm carbon fiber [
]. The distance between the detector plane and the top of the support paddle was 1.5 cm. The breast in CC view was modelled as a cylinder with a radius of 8.0 cm. In order to explore the impact of simulating the compressed breast in MLO view in mammographic examinations, we replicated the model proposed in Ref. [
]. Here, the breast model included also a portion of the pectoralis muscle, as visible in a mammographic MLO view. The simulation parameters included the glandular fraction by mass, the breast thickness and the chest-to-nipple distance (CND). Fig. 2 shows the scheme of the two compressed breast models. As in Ref. [
] for the DBT exams, we varied the CND value in the range 7–13 cm in order to explore its impact on the MGD estimates in planar mammography. Furthermore, in DBT CC views we investigated the relationship between T and t factors and beam HVL, breast characteristics and scanning geometry. For the MGD calculation in mammography, we adopted the Mo/Mo spectrum at 30 kV reported in the TG195 report [
In order to simulate the BCT irradiation geometry for the calculation of DgNCT, the pendant breast was modelled as a vertical cylinder enveloped in a 1.45-mm thick layer simulating the skin. The breast was characterized by its diameter, length and glandular fraction by mass. In a previous paper [
] we calculated monoenergetic DgNCT coefficients in the energy range 8–80 keV and polyenergetic DgNCT coefficients for 49 kV spectra (for 5 HVL levels). In that paper, the conversion coefficients were calculated for a rectangular scoring surface of 18 × 30 mm2 for determination of the KISO, which reflected the cross section of the 6 cm3 ion chamber that we used for KISO measurements. In this work, however, we considered the use of a 100-mm long CT ion chamber for KISO measurements: hence, here we calculated the ion-chamber conversion coefficient, IC, in order to convert those previously-determined DgNCT coefficients to the DgNCT usable with the new type of ion chamber (DgNCT100mm):
(10)
The provision of this IC coefficient allowed us to investigate the influence of the area of the scoring surface on the DgNCT values calculated with our MC code. KISO6 is the simulated air kerma for a 18 × 30 mm2 photon fluence scoring surface which reflects the cross section of 6 cm3 ion chamber and KISO10 is the simulated air kerma for a 6 × 100 mm2 scoring area for the photon fluence, which reflects the cross section of a 3 cm3 ion chamber. The 30-mm long and the 100-mm long sides are parallel to the rotation axis of the simulated BCT scanner. Both scoring surfaces were placed at the simulated scanner isocenter. DgNCT100mm was finally computed as follows:
Fig. 3 shows the simulated MGD per photon, in a CC view and in a MLO view in 2D mammography. The CND for MLO view was set in order to produce a breast model with the same glandular mass of the CC view model (with a diameter of 8 cm). The distances between the source and the detector and between the source and the top of the compressed breast were in both cases 660 mm and 595 mm, respectively, with an air gap of 15 mm between the bottom surface of the breast and the detector plane. Simulations were run for a 30 kV Mo/Mo spectrum. The difference in MGD for CC and for MLO views was less than 0.4%.
Fig. 3MGD per photon for a 20% glandular breast compressed to 50 mm in CC view and MLO view. The CND in MLO view was tuned in order to have a breast with the same glandular mass in both models. Skin layer = 1.45 mm skin tissue.
Fig. 4 shows the simulated MGD per photon for a breast in MLO view, for various CND values in the range 7–13 cm. The MGD is almost constant in the range 7–9 cm of CND, and it decreases for larger CND values. The maximum difference in the values of MGD per photon is 1.4% of the maximum value.
Fig. 4MGD per photon in MLO view as a function of chest-to-nipple distance. Skin layer = 1.45 mm skin tissue.
]. The simulated compressed breast was 50-mm thick with a glandular fraction by mass of 50%. In all cases reported in Fig. 5, the relative differences in t-factor values with respect to the data in Ref. [
Fig. 6 reports t-factors as a function of projection angle, for four compressed breast thicknesses in the range 20–110 mm. For scan angles less than 15 degrees, the thicker is the breast, the lower is the corresponding t-factor. For greater scan angles, no monotonic dependence between t-factors and breast thickness was observed. For the thickest breasts (80 mm and 110 mm) and for scan angles lower than 20 degrees, t-factor differences are negligible; for higher scan degrees, t-factors for 110-mm thick breast resulted higher than those for 80-mm thick breasts. Modelling the skin envelope as a 1.45-mm thick skin tissue layer (instead of 5 mm adipose tissue) increases the t-factors (Fig. 7, 34 kV W/Rh and 29 kV Rh/Rh spectra, 50 mm breast thickness). The differences are greater for increasing scan angles. For instance, for 29 kV Rh/Rh spectrum, for a 15 degrees scan angle the former model produced a t-factor 0.5% larger than the one obtained with a 5 mm adipose envelope; this relative difference increased to 3.3% for a projection angle of 30 degrees.
Fig. 6t-factors for various compressed breast thicknesses. SID = 620 mm; IDD = 40 mm. Skin layer = 5 mm adipose tissue.
For fixed tube voltage and anode filter combination, the HVL has little influence on the T-factor (Fig. 8). For 29 kV Rh/Rh spectra and a 30 mm compressed breast, using a spectrum with an HVL of 0.410 mmAl instead of 0.337 mmAl increased the T-factor by 0.8% (Fig. 8a). The maximum deviation among the T-factors computed for 34 W/Rh spectra and 50 mm breast thickness does not exceed 0.3% (Fig. 8b). Fig. 9 shows the T-factors at various source-to-isocenter distances. For fixed breast characteristics and source-to-detector distance, the T-factor decreased by increasing the SID; increasing the SID from 55 cm to 65 cm reduced the T-factor by 1.2%.
Fig. 8T-factors at varying HVL beam values for a) 30 mm breast thickness and 29 kV Rh/Rh beam spectrum and b) 50 mm breast thickness and 34 kV Rh/Rh beam spectrum. SID = 650 mm; IDD = 47 mm. 25 projections over 50 degrees scan angle. Skin layer = 5 mm adipose tissue.
Fig. 9T-factors at source-to-isocenter distances. The source to isocenter distance was fixed at 697 mm. 25 projections over 50 degrees scan angle. Skin layer = 5 mm adipose tissue.
]. These last are computed for a 6 cm3 ion chamber (for air kerma determinations) with a cross section of 18 × 30 mm2. IC coefficients were computed for photon energies between 8 keV and 80 keV. Since the difference is due to the sole geometrical configuration of the scoring region, the coefficients are not expected to depend on the X-ray beam energy. The average IC value was 1.01, indicating that there is a 1% increase in the estimate of KISO when changing the scoring area from 6 × 100 mm2 to 18 × 30 mm2, an effect due to the divergence of the X-ray beam and consequent variation of the air kerma in different positions in the isocenter irradiation plane. Due to its independence on the beam energy, this value can be also used for polyenergetic DgNCT coefficients, when changing the KISO scoring surface area between the two values (18 × 30 mm2 and 6 × 100 mm2).
Fig. 10Conversion factor between DgNCT calculated for a 6 cm3 ion chamber and the one calculated for air kerma measured by means of a 100-mm CT ion chamber.
Fig. 11 shows pDgNCT values for a 80 kV W/Cu spectrum. They are calculated for a cylindrical breast model with a length equal to 1.5 times its radius. Data in Fig. 11, together with those calculated for a breast with length of 1 or 2 times the breast radius, are reported in the Supplementary material. Data reported in Ref. [
] (calculated for 50% glandular breast) are also reported in Fig. 11, for comparison. The latter data are 7% higher for the case of an 8-cm breast diameter and the difference decreases as the breast diameter increases, down to a difference of 2% for a 18-cm breast diameter (Fig. 11).
In order to contribute to the process of reviewing the paradigm in X-ray breast imaging dosimetry, we developed a Monte Carlo code for assessing normalized glandular dose coefficients in 2D and 3D breast dosimetry. In this paper, we presented the most recent version of this code, including the added features for computing the MGD in DBT and in MLO views. We replicated the compressed breast model proposed in Ref. [
] for representing the breast in MLO mammographic view, which comprises a portion of the pectoral muscle. Using the MC estimates, we provided indication that adopting a suitable breast model for such a view does not produce large differences in MGD compared to that obtained for a conventional compressed breast model adopted for representing the organ in CC view. The difference for the simulated Mo/Mo spectrum at 30 kV, and a 50 mm thick and 50% glandular breast, was about 0.4%. As already shown in mammography [
], the specific skin model adopted may have a significant impact on MGD estimates in DBT. Here we compared, for two X-ray spectra, the differences in the computed t-factors by adopting a 1.45 mm of skin layer instead of a 5 mm adipose layer, as envisaged in current quality assurance protocols. In the performed simulations, a 1.45-mm thick skin layer produced higher t-factors, for any scan angle in the range 0–30 degrees. However, the larger is the scan angle, the larger is the influence of the skin model on MGD estimates. For a 50 mm breast thickness and 50% glandular fraction by mass, and for a 29 kV Rh/Rh spectra and 15 degrees scan angle, a 1.45-mm thick skin layer led to t-value 0.5% higher than that for 5 mm adipose layer; such a difference increased to 3.3% for a scan angle of 30 degrees in a DBT scan. When all the other parameter are fixed, varying the beam HVL did not produce any significant difference in T-factors.
], we calculated the DgNCT for MGD estimates in BCT. There, a 18 × 30 mm2 scoring surface for the air kerma calculation was adopted: this choice pointed to using an ion chamber of 6 cm3 sensitive volume at the scanner isocenter, for measuring the air kerma (as required in Eq. (3)). On the other hand, when using a 100 mm CT ion chamber for such a measurement, the scoring surface changes (to 6 × 100 mm2). In order to calculate the corresponding new DgNCT coefficients, we have calculated an (energy independent) correction factor IC which is the ratio of the two simulated KISO values. This IC ratio is expected to be slightly greater than 1 since the 100-mm long ion chamber extends at the axis of rotation at larger cone angles than the 6-cm3 chamber, where the air kerma decreases due to beam divergence. Its value (IC = 1.01 as calculated with our MC code) can be used for the case of either monoenergetic or polyenergetic X-ray beams in order to convert DgNCT values obtained with either ion chambers for KISO measurements. Polyenergetic DgNCT were calculated for an 80 kV spectrum (adopted in Ref. [
]) and are reported in the Supplementary material in tabular form. The DgNCT values calculated with the MC code in this work showed differences between 2% and 7% from the values reported in Ref. [
We presented the most recent version of our Monte Carlo code for the MGD calculation in 2D and 3D X-ray breast imaging. It comprises also a module for the t-factors and T-factors calculation in DBT. The code was validated vs. literature data reported in Ref. [
]. We developed a compressed breast model for representing the breast undergoing a mammographic exam in MLO view. The MGD calculated with such a model did not significantly differ from that calculated for the model of the breast undergoing a CC view exam. The skin model in DBT presented a significant influence on the calculated t-factor. A breast with a 1.45 mm skin produced higher t-factors than that with a skin simulated as a 5 mm adipose tissue, at any scan angle. The larger is the scan angle, the larger is the difference. Finally, we computed a conversion factor for the use of DgNCT coefficients reported in Ref. [
] in conjunction with air kerma measurements by means of 100 mm CT ion chamber. Polyenergetic DgNCT coefficients were calculated with the new breast model, for the 80 kV spectrum adopted by a clinical BCT scanner developed at UC Davis.
Acknowledgements
The authors would like to thank prof. D. R. Dance for sharing data of the Ref. [
] reported in Fig. 5. This work was supported by INFN and by the MAXIMA project: Three dimensional breast cancer models for X-ray imaging research. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 692097.
Institute of Physics and Engineering in Medicine (IPEM). The commissioning and routine testing of mammographic X-ray systems. IPEM Report 89 (York, United Kingdom: IPEM); 2005.
Estimation of mean glandular dose for contrast enhanced digital mammography: factors for use with the UK, European and IAEA breast dosimetry protocols.
31163-3&id=gr1.jpg){kind=link}
31163-3&id=gr2.jpg){kind=link}
31163-3&id=gr3.jpg){kind=link}
31163-3&id=gr4.jpg){kind=link}
31163-3&id=gr5.jpg){kind=link}
31163-3&id=gr6.jpg){kind=link}
31163-3&id=gr7.jpg){kind=link}
31163-3&id=gr8.jpg){kind=link}
31163-3&id=gr9.jpg){kind=link}
31163-3&id=gr10.jpg){kind=link}
31163-3&id=gr11.jpg){kind=link}