Multiscale Monte Carlo simulations for dosimetry in x-ray breast imaging: Part II - Microscopic scales.

BACKGROUND: Although the benefits of breast screening and early diagnosis are known for reducing breast cancer mortality rates, the effects and risks of low radiation doses to the cells in the breast are still ongoing topics of study. PURPOSE: To study specific energy distributions ( f ( z , D g ) $f(z,D_{g})$ ) in cytoplasm and nuclei of cells corresponding to glandular tissue for different x-ray breast imaging modalities. METHODS: A cubic lattice (500 µm length side) containing 4064 spherical cells was irradiated with photons loaded from phase space files with varying glandular voxel doses ( D g $D_{g}$ ). Specific energy distributions were scored for nucleus and cytoplasm compartments using the PENELOPE (v. 2018) + penEasy (v. 2020) Monte Carlo (MC) code. The phase space files, generated in part I of this work, were obtained from MC simulations in a voxelized anthropomorphic phantom corresponding to glandular voxels for different breast imaging modalities, including digital mammography (DM), digital breast tomosynthesis (DBT), contrast enhanced digital mammography (CEDM) and breast CT (BCT). RESULTS: In general, the average specific energy in nuclei is higher than the respective glandular dose scored in the same region, by up to 10%. The specific energy distributions for nucleus and cytoplasm are directly related to the magnitude of the glandular dose in the voxel ( D g $D_{g}$ ), with little dependence on the spatial location. For similar D g $D_{g}$ values, f ( z , D g ) $f(z,D_{g})$ for nuclei is different between DM/DBT and CEDM/BCT, indicating that distinct x-ray spectra play significant roles in f ( z , D g ) $f(z,D_{g})$ . In addition, this behavior is also present when the specific energy distribution ( F g ( z ) $F_{g}(z)$ ) is considered taking into account the GDD in the breast. CONCLUSIONS: Microdosimetry studies are complementary to the traditional macroscopic breast dosimetry based on the mean glandular dose (MGD). For the same MGD, the specific energy distribution in glandular tissue varies between breast imaging modalities, indicating that this effect could be considered for studying the risks of exposing the breast to ionizing radiation.

Multiscale Monte Carlo simulations for dosimetry in x-ray breast imaging: Part I - Macroscopic scales.

BACKGROUND: X-ray breast imaging modalities are commonly employed for breast cancer detection, from screening programs to diagnosis. Thus, dosimetry studies are important for quality control and risk estimation since ionizing radiation is used. PURPOSE: To perform multiscale dosimetry assessments for different breast imaging modalities and for a variety of breast sizes and compositions. The first part of our study is focused on macroscopic scales (down to millimeters). METHODS: Nine anthropomorphic breast phantoms with a voxel resolution of 0.5 mm were computationally generated using the BreastPhantom software, representing three breast sizes with three distinct values of volume glandular fraction (VGF) for each size. Four breast imaging modalities were studied: digital mammography (DM), contrast-enhanced digital mammography (CEDM), digital breast tomosynthesis (DBT) and dedicated breast computed tomography (BCT). Additionally, the impact of tissue elemental compositions from two databases were compared. Monte Carlo (MC) simulations were performed with the MC-GPU code to obtain the 3D glandular dose distribution (GDD) for each case considered with the mean glandular dose (MGD) fixed at 4 mGy (to facilitate comparisons). RESULTS: The GDD within the breast is more uniform for CEDM and BCT compared to DM and DBT. For large breasts and high VGF, the ratio between the minimum/maximum glandular dose to MGD is 0.12/4.02 for DM and 0.46/1.77 for BCT; the corresponding results for a small breast and low VGF are 0.35/1.98 (DM) and 0.63/1.42 (BCT). The elemental compositions of skin, adipose and glandular tissue have a considerable impact on the MGD, with variations up to 30% compared to the baseline. The inclusion of tissues other than glandular and adipose within the breast has a minor impact on MGD, with differences below 2%. Variations in the final compressed breast thickness alter the shape of the GDD, with a higher compression resulting in a more uniform GDD. CONCLUSIONS: For a constant MGD, the GDD varies with imaging modality and breast compression. Elemental tissue compositions are an important factor for obtaining MGD values, being a source of systematic uncertainties in MC simulations and, consequently, in breast dosimetry.

Diagnostic reference level quantities for adult chest and abdomen-pelvis CT examinations: correlation with organ doses.

OBJECTIVES: To evaluate correlations between DRL quantities (DRLq) stratified into patient size groups for non-contrast chest and abdomen-pelvis CT examinations in adult patients and the corresponding organ doses. METHODS: This study presents correlations between DRLq  (CTDIvol, DLP and SSDE) stratified into patient size ranges and corresponding organ doses shared in four groups: inside, peripheral, distributed and outside. The demographic, technical and dosimetric parameters were used to identify the influence of these quantities in organ doses. A robust statistical method was implemented in order to establish these correlations and its statistical significance. RESULTS: Median values of the grouped organ doses are presented according to the effective diameter ranges. Organ doses in the regions inside the imaged area are higher than the organ doses in peripheral, distributed and outside regions, excepted to the peripheral doses associated with chest examinations. Different levels of statistical significance between organ doses and the DRLq were presented. CONCLUSIONS: Correlations between DRLq and target-organ doses associated with clinical practice can support guidance's to the establishment of optimization criteria. SSDE demonstrated to be significant in the evaluation of organ doses is also highlighted. The proposed model allows the design of optimization actions with specific risk-reduction results.

Impact of fibroglandular tissue distribution and breast shape in voxelized breast models for dosimetry in mammography.

Objective.This work proposes to study the impact of different voxelized heterogeneous breast models (gaussian centered - GaussC; gaussian lower - GaussL; and fitted equation patient-based on 3D realistic distribution (Fedonet al2021) - FitPB) for dosimetry in mammography compared to a well-established homogeneous approximation. Influence of breast outer shape also was investigated by comparing semicylindric and anthropomorphic breasts.Approach.By using the PENELOPE (v. 2018) + penEasy (v. 2020) MC code, simulations were performed to evaluate the normalized glandular dose (DgN) and the glandular depth dose (GDD(z)) for different breast characteristics and x-ray beam spectra.Main results.The averageDgNoverestimation caused by homogeneous tissue approximation was 33.0%, with the highest values attributed to GaussLand FitPBmodels, where fibroglandular tissue is concentrated deeper in the breast. The observed variation between anthropomorphic and semicylindrical breast shapes was, on average, 5.6%, legitimizing the latter approximation for breast dosimetry. Thicker breasts and lower energy beams resulted in larger overestimation caused by the homogeneous approach, while variations inDgNvalues among different heterogeneous models were higher for thinner breast and lower energy beams. Moreover, the depth where differences betweenGDD(z) for different breast models became maximum depends on the axial variation of fibroglandular tissue concentration between each model. TheGDD(z) dependence results in a significant variation of the contribution of each breast depth to mean glandular dose (MGD) among the breast models studied.Significance.Intercomparison between different breast models for dosimetry can be useful for estimating more accurateMGDvalues for population-based dosimetry, for exploring the use of 1D gaussian distribution for breast dosimetry, and for understanding the dose distributions inside the fibroglandular tissues, which could be a novel source of information for risk estimations.

Technical note: MC-GPU breast dosimetry validations with other Monte Carlo codes and phase space file implementation.

PURPOSE: To validate the MC-GPU Monte Carlo (MC) code for dosimetric studies in X-ray breast imaging modalities: mammography, digital breast tomosynthesis, contrast enhanced digital mammography, and breast-CT. Moreover, to implement and validate a phase space file generation routine. METHODS: The MC-GPU code (v. 1.5 DBT) was modified in order to generate phase space files and to be compatible with PENELOPE v. 2018 derived cross-section database. Simulations were performed with homogeneous and anthropomorphic breast phantoms for different breast imaging techniques. The glandular dose was computed for each case and compared with results from the PENELOPE (v. 2014) + penEasy (v. 2015) and egs _ brachy (EGSnrc) MC codes. Afterward, several phase space files were generated with MC-GPU and the scored photon spectra were compared between the codes. The phase space files generated in MC-GPU were used in PENELOPE and EGSnrc to calculate the glandular dose, and compared with the original dose scored in MC-GPU. RESULTS: MC-GPU showed good agreement with the other codes when calculating the glandular dose distribution for mammography, mean glandular dose for digital breast tomosynthesis, and normalized glandular dose for breast-CT. The latter case showed average/maximum relative differences of 2.3%/27%, respectively, compared to other literature works, with the larger differences observed at low energies (around 10 keV). The recorded photon spectra entering a voxel were similar (within statistical uncertainties) between the three MC codes. Finally, the reconstructed glandular dose in a voxel from a phase space file differs by less than 0.65%, with an average of 0.18%-0.22% between the different MC codes, agreement within approximately 2 σ statistical uncertainties. In some scenarios, the simulations performed in MC-GPU were from 20 up to 40 times faster than those performed by PENELOPE. CONCLUSIONS: The results indicate that MC-GPU code is suitable for breast dosimetric studies for different X-ray breast imaging modalities, with the advantage of a high performance derived from GPUs. The phase space file implementation was validated and is compatible with the IAEA standard, allowing multiscale MC simulations with a combination of CPU and GPU codes.

Breast glandularity and mean glandular dose assessment using a deep learning framework: Virtual patients study.

PURPOSE: Breast dosimetry in mammography is an important aspect of radioprotection since women are exposed periodically to ionizing radiation due to breast cancer screening programs. Mean glandular dose (MGD) is the standard quantity employed for the establishment of dose reference levels in retrospective population studies. However, MGD calculations requires breast glandularity estimation. This work proposes a deep learning framework for volume glandular fraction (VGF) estimations based on mammography images, which in turn are converted to glandularity values for MGD calculations. METHODS: 208 virtual breast phantoms were generated and compressed computationally. The mammography images were obtained with Monte Carlo simulations (MC-GPU code) and a ray-tracing algorithm was employed for labeling the training data. The architectures of the neural networks are based on the XNet and multilayer perceptron, adapted for each task. The network predictions were compared with the ground truth using the coefficient of determination (r2). RESULTS: The results have shown a good agreement for inner breast segmentation (r2 = 0.999), breast volume prediction (r2 = 0.982) and VGF prediction (r2 = 0.935). Moreover, the DgN coefficients using the predicted VGF for the virtual population differ on average 1.3% from the ground truth values. Afterwards with the obtained DgN coefficients, the MGD values were estimated from exposure factors extracted from the DICOM header of a clinical cohort, with median(75 percentile) values of 1.91(2.45) mGy. CONCLUSION: We successfully implemented a deep learning framework for VGF and MGD calculations for virtual breast phantoms.

Impact of photoelectric cross section data on systematic uncertainties for Monte Carlo breast dosimetry in mammography.

Monte Carlo (MC) simulations are employed extensively in breast dosimetry studies. In the energy interval of interest in mammography energy deposition is predominantly caused by the photoelectric effect, and the corresponding cross sections used by the MC codes to model this interaction process have a direct influence on the simulation results. The present work compares two photoelectric cross section databases in order to estimate the systematic uncertainty, related to breast dosimetry, introduced by the choice of cross sections for photoabsorption. The databases with and without the so-called normalization screening correction are denoted as 'renormalized' or 'un-normalized', respectively. The simulations were performed with the PENELOPE/penEasy code system, for a geometry resembling a mammography examination. The mean glandular dose (MGD), incident air kerma (Kair), normalized glandular dose (DgN) and glandular depth-dose (GDD(z)) were scored, for homogeneous breast phantoms, using both databases. The AAPM Report TG-195 case 3 was replicated, and the results were included. Moreover, cases with heterogeneous and anthropomorphic breast phantoms were also addressed. The results simulated with the un-normalized cross sections are in better overall agreement with the TG-195 data than those from the renormalized cross sections; for MGD the largest discrepancies are 0.13(6)% and 0.74(5)%, respectively. The MGD,Kairand DgN values simulated with the two databases show differences that diminish from approximately 10%/3%/6.8% at 8.25 keV down to 1.5%/1.7%/0.4% at 48.75 keV, respectively. For polyenergetic spectra, deviations up to 2.5% were observed. The disagreement between the GDDs simulated with the analyzed databases increases with depth, ranging from -1% near the breast entrance to 4% near the bottom. Thus, the choice of photoelectric cross section database affects the MC simulation results of breast dosimetry and adds a non-negligible systematic uncertainty to the dosimetric quantities used in mammography.

Estimation of glandular dose in mammography based on artificial neural networks.

This work proposes using artificial neural networks (ANNs) for the regression of the dosimetric quantities employed in mammography. The data were generated by Monte Carlo (MC) simulations using a modified and validated version of the PENELOPE (v. 2014) + penEasy (v. 2015) code. A breast model of a homogeneous mixture of adipose and glandular tissue was adopted. The ANNs were constructed using the Keras and scikit-learn libraries for mean glandular dose (MGD) and air kerma (Kair ) regressions, respectively. In total, seven parameters were considered, including the incident photon energies (from 8.25 to 48.75 keV), breast geometry, breast glandularity and Kair acquisition geometry. Two ensembles of five ANNs each were formed to calculate MGD and Kair . The normalized glandular dose coefficients (DgN) were calculated using the ratio of the ensemble outputs for MGD and Kair . Polyenergetic DgN values were calculated by weighting monoenergetic values by the spectrum bin probabilities. The results indicate a very good ANN prediction performance when compared to the validation data, with median errors on the order of the average simulation uncertainties (≈ 0.2%). Moreover, the predicted DgN values are in good agreement compared with previously published works, with mean (maximum) differences up to 2.2% (9.4%). Therefore, it is shown that ANNs could be a complementary or alternative technique to tables, parametric equations and polynomial fits to estimate DgN values obtained via MC simulations.

Normalized glandular dose (DgN) coefficients from experimental mammographic x-ray spectra.

Mean glandular dose is the quantity used for dosimetry in mammography and depends on breast-related characteristics, such as thickness and density, and on the x-ray spectrum used for breast imaging. This work aims to present an experimentally-based method to derive polyenergetic normalized glandular dose coefficients (DgNp) from the spectral difference between x-ray spectra incident and transmitted through breast phantoms with glandular/adipose proportions of 30/70 and 50/50 and thicknesses up to 4.5 cm. The spectra were produced by a Mammomat 3000 Nova system using radiographic techniques commonly applied for imaging compressed breast thickness lower than 6 cm (Mo/Mo, Mo/Rh and W/Rh spectra at 26 and 28 kVp). DgNp coefficients were compared with values estimated using Boones' method and data from breast images (DICOM Organ Dose and VolparaDose calculations). The DgNp were also evaluated in layers into the phantoms (depth-DgNp) using both x-ray spectra and thermoluminescent dosimeters (TLD-100). Maximum differences between DgNp from the method presented in this study and results using Boone's method was 11%, with larger differences for Mo/Rh spectra in relation to the Mo/Mo. The DgNp maximum differences to the coefficients obtained using patient images were 8.0%, for the DgN calculated using Volpara and 6.4% for the DgN from DICOM Organ Dose, for a 4.5 cm breast phantom with 30% glandularity. The DgNp estimated from the depth-DgNp distributions differ up to 5.2% to the coefficients obtained using the pair incident-transmitted spectra to calculate the DgNp directly in the whole phantom. The depth-DgNp distributions estimated with TLDs were consistent with the results observed using the experimental spectra, with maximum difference of 3.9%. In conclusion, polyenergetic x-ray spectrometry proved to be an applicable tool for research in dosimetry in mammography allowing spectral characterization. This approach can also be useful for investigation of the influence of x-ray spectra on glandular dose.

Skin models and their impact on mean glandular dose in mammography.

PURPOSE: To quantify the influence of different skin models on mammographic breast dosimetry, based on dosimetric protocols and recent breast skin thickness findings. METHODS: By using an adapted PENELOPE (v. 2014) + PenEasy (v. 2015) Monte Carlo (MC) code, simulations were performed in order to obtain the mean glandular dose (MGD), the normalized MGD by incident air Kerma (DgN), and the glandular depth dose (GDD(z)). The geometry was based on a cranio-caudal mammographic examination. Monoenergetic and polyenergetic beams were implemented, for a breast thickness from 2 cm to 9 cm, with different compositions. Seven skin models were used: a 5 mm adipose layer; a skin layer ranging from 5 mm to 1.45 mm, a 1.45 mm skin thickness with a subcutaneous adipose layer of 2 mm and 3.55 mm. RESULTS: The differences, for monoenergetic beams, are higher (up to 200%) for lower energies (8 keV), thicker and low glandular content breasts, decreasing to less than 5% at 40 keV. Without a skin layer, the differences reach a maximum of 1240%. The relative difference in DgN values for 1.45 mm skin and 5 mm adipose layers and polyenergetic beams varies from -14% to 12%. CONCLUSIONS: The implemented MC code is suitable for mammography dosimetry calculations. The skin models have major impacts on MGD values, and the results complement previous literature findings. The current protocols should be updated to include a more realistic skin model, which provides a reliable breast dose estimation.

Direct measurement of clinical mammographic x-ray spectra using a CdTe spectrometer.

PURPOSE: To introduce and evaluate a method developed for the direct measurement of mammographic x-ray spectra using a CdTe spectrometer. The assembly of a positioning system and the design of a simple and customized alignment device for this application is described. METHODS: A positioning system was developed to easily and accurately locate the CdTe detector in the x-ray beam. Additionally, an alignment device to line up the detector with the central axis of the radiation beam was designed. Direct x-ray spectra measurements were performed in two different clinical mammography units and the measured x-ray spectra were compared with computer-generated spectra. In addition, the spectrometer misalignment effect was evaluated by comparing the measured spectra when this device is aligned relatively to when it is misaligned. RESULTS: The positioning and alignment of the spectrometer have allowed the measurements of direct mammographic x-ray spectra in agreement with computer-generated spectra. The most accurate x-ray spectral shape, related with the minimal HVL value, and high photon fluence for measured spectra was found with the spectrometer aligned according to the proposed method. The HVL values derived from both simulated and measured x-ray spectra differ at most 1.3 and 4.5% for two mammography devices evaluated in this study. CONCLUSION: The experimental method developed in this work allows simple positioning and alignment of a spectrometer for x-ray spectra measurements given the geometrical constraints and maintenance of the original configurations of mammography machines.

Application of a semi-empirical model for the evaluation of transmission properties of barite mortar.

The aim of this study was to estimate barite mortar attenuation curves using X-ray spectra weighted by a workload distribution. A semi-empirical model was used for the evaluation of transmission properties of this material. Since ambient dose equivalent, H(⁎)(10), is the radiation quantity adopted by IAEA for dose assessment, the variation of the H(⁎)(10) as a function of barite mortar thickness was calculated using primary experimental spectra. A CdTe detector was used for the measurement of these spectra. The resulting spectra were adopted for estimating the optimized thickness of protective barrier needed for shielding an area in an X-ray imaging facility.