El futuro es Transmedia

Further to the cross section validation, the MC code has been macroscopically validated using mammography geometries described in prior published work. Three different experiments published by Boone et al. [61], using a cellular anti-scatter grid described by Rezentes et al. [58], have been simulated.

The cellular anti-scatter grid has been simulated in GEANT4 using the appropriate dimensions for each experiment. Its design consists of a rectangular box filled in with the septa material (e.g. copper). Then, a series of parallelepipeds (aligned with the focal spot) filled with the appropriate material (e.g. air) are placed within the rectangular box, simulating the interspace.

These publications were chosen because they shows results for a cellular anti-scatter grid design (Lorad M-III) and it will validate the methodology used to implement the

4.5. Validation of simulations 89

Figure 4.11: Comparison between GEANT4 cross sections (blue squares) and NIST-XCOM cross sections (red circles) for cesium. Incoherent, coher- ent and photoelectric physical interactions are shown for an energy range from 1keV to 50 keV.

Hologic Selenia Dimensions anti-scatter grid (also cellular) in future simulations. The dimensions and material of this grid are shown in Table 4.6.

For each experiment, five simulations using 109 photons were used. The resulting images were median filtered using a 11x11 kernel and a pixel size of 1mm2 was used. The SEM associated to each of the calculated metrics is shown in the results below. The MC setup used in these experiments is shown in Figure 4.3(b). A semicircular breast phantom with 10cm radius was used. The phantom composition was 50% adipose 50% glandular breast tissue and neither compression paddle nor breast support were included in the simulations described by Boone et al. [61]. The specific configurations used for each experiment are summarised in Table 4.7.

For all the experiments, the CIF and BF were calculated using methodology explained in Section 2.4.5. The primary plus scatter image (P+S) and the SPR were calculated with (G) and without (NG) the anti-scatter grid in a 50x90 pixels ROI (see red square in Figure 4.12(a)) using a pixel size of 1mm2_.

The primary, scatter and SPR images recorded for Experiment I with and without the anti-scatter grid are shown in Figure 4.12.

The results from experiments I, II and III are shown in Tables 4.8, 4.9 and 4.10 respectively. Overall, it has been found less than 3% discrepancy in CIF and BF with Boone et al. data [61], the error being attributed to different HVL values employed as the author (J.M. Boone) did not specified them in his paper. Furthermore, both the statistical uncertainties and cross sections of his MC results are not described in the paper. Results from experiment II shows a discrepancy of up to approximately 8%

90 Chapter 4. Monte Carlo simulations of X-ray mammography

Table 4.5: Quantitative comparison between incoherent, coherent and pho- toelectric cross sections for cesium measured from GEANT4 and NIST- XCOM for an energy range between 1keV to 50keV. Note that the largest discrepancies are highlighted in red.

Incoherent Coherent Photoelectric

Energy GEANT4 NIST-XCOM Difference GEANT4 NIST-XCOM Difference GEANT4 NIST-XCOM Difference

(keV) (cm2_{/g) (cm}2_{/g) (%) (cm}2_{/g) (cm}2_{/g) (%) (cm}2_{/g) (cm}2_{/g) (%)} 1 5.89x10−3 _5.88x10−3 _{0.1 5.73x10}0 _8.55x100 _{33.0 9.45x10}3 _9.36x103 _1.0 1.032 6.13x10−3 _6.13x10−3 _{0.0 5.76x10}0 _8.52x100 _{32.5 8.83x10}3 _8.76x103 _0.8 1.065 6.39x10−3 _6.39x10−3 _{0.1 6.07x10}0 _8.49x100 _{28.5 8.51x10}3 _8.21x103 _3.7 1.065 6.39x10−3 _6.39x10−3 _{0.1 6.07x10}0 _8.49x100 _{28.5 8.51x10}3 _8.68x103 _2.0 1.139 6.97x10−3 _6.93x10−3 _{0.1 6.44x10}0 _8.42x100 _{23.5 7.59x10}3 _7.56x103 _0.5 1.217 7.59x10−3 _7.59x10−3 _{0.1 6.71x10}0 _8.35x100 _{19.6 6.87x10}3 _6.58x103 _4.5 1.217 7.59x10−3 _7.59x10−3 _{0.1 5.71x10}0 _8.35x100 _{19.6 6.87x10}3 _6.88x103 _0.1 1.5 9.88x10−3 _9.86x10−3 _{0.1 7.11x10}0 _8.07x100 _{11.9 4.33x10}3 _4.33x103 _0.1 2 1.40x10−2 _1.40x10−2 _{0.2 6.78x10}0 _7.58x100 _{10.5 2.22x10}3 _2.22x103 _0.0 3 2.21x10−2 _2.20x10−2 _{0.2 5.59x10}0 _6.59x100 _{15.2 8.25x10}2 _8.25x102 _0.0 4 2.95x10−2 _2.94x10−2 _{0.3 4.40x10}0 _5.75x100 _{23.5 3.99x10}2 _4.00x102 _0.1 5 3.60x10−2 _3.58x10−2 _{0.4 2.48x10}0 _5.04x100 _{50.8 2.25x10}2 _2.25x102 _0.1 5.012 3.60x10−2 _3.59x10−2 _{0.4 2.50x10}0 _5.03x100 _{50.4 6.66x10}2 _2.24x102 _197.5 5.012 3.60x10−2 _3.59x10−2 _{0.4 2.50x10}0 _5.03x100 _{50.4 6.66x10}2 _6.63x102 _0.6 5.183 3.70x10−2 _3.69x10−2 _{0.3 2.99x10}0 _4.92x100 _{39.1 6.14x10}2 _6.09x102 _0.8 5.359 3.80x10−2 _3.79x10−2 _{0.2 3.02x10}0 _4.82x100 _{37.4 5.63x10}2 _5.60x102 _0.6 5.359 3.80x10−2 _3.79x10−2 _{0.2 3.02x10}0 _4.82x100 _{37.4 5.63x10}2 _7.64x102 _26.4 5.534 3.89x10−2 _3.89x10−2 _{0.2 3.39x10}0 _4.72x100 _{28.2 7.09x10}2 _7.05x102 _0.6 5.714 3.99x10−2 3.99x10−2 0.2 3.54x100 _4.62x100 _{23.4 7.52x10}2 _6.51x102 _15.5 5.714 3.99x10−2 3.99x10−2 0.2 3.54x100 _4.62x100 _{23.4 7.52x10}2 _7.50x102 _0.3 6 4.15x10−2 4.14x10−2 0.3 3.88x100 _4.46x100 _{13.0 6.66x10}2 _6.67x102 _0.1 8 5.11x10−2 5.08x10−2 0.6 3.60x100 _3.55x100 _{1.5 3.17x10}2 _3.18x102 _0.2 10 5.93x10−2 5.89x10−2 0.6 2.97x100 _2.89x100 _{3.0 1.76x10}2 _1.76x102 _0.1 15 7.51x10−2 7.46x10−2 0.6 1.91x100 _1.90x100 _{0.8 5.90x10}1 _5.91x101 _0.1 20 8.59x10−2 8.52x10−2 0.8 1.32x100 _1.36x100 _{2.6 2.67x10}1 _2.68x101 _0.1 30 9.83x10−2 9.76x10−2 0.8 6.95x10−1 7.79x10−1 10.8 8.62x100 _8.63x100 _0.1 35.98 1.03x10−1 1.02x10−1 0.7 4.17x10−1 5.95x10−1 29.9 5.16x100 _5.17x100 _0.1 35.98 1.03x10−1 1.02x10−1 0.7 4.17x10−1 5.95x10−1 29.9 5.16x100 _3.07x101 _83.2 40 1.05x10−1 1.04x10−1 0.9 4.65x10−1 5.07x10−1 8.3 2.33x101 _2.32x101 _0.6 50 1.09x10−1 1.08x10−1 1.0 3.58x10−1 3.61x10−1 0.8 1.29x101 _1.29x101 _0.2

Table 4.6: Specifications of Lorad M-III cellular grid described by Rezentes et al. [58].

Grid Grid Interspace Septa Septa Septa Interspace type ratio(h/D) material material height (h) thickness (d) thickness (D) Cellular 3.8 Air Copper 2.4mm 0.03mm 0.64mm

Table 4.7: Configuration used for each experiment described by Boone et al. [61].

Experiment X-ray spectrum Breast thickness Detector (thickness/composition) I 28kVp Mo/Mo 6cm Kodak Min-R (46µm Gd2O2S)

II 40kVp W/Al 6cm Ideal detector

III 30kVp Mo/Mo 8cm Kodak Min-R (46µm Gd2O2S)

in SPR values. It is not stated in the paper, but in the past, it has been found that this author (J.M. Boone) uses slightly different glandular and adipose tissue composi- tion. This might show larger discrepancies for this case, where the maximum energy simulated was 40kVp.

4.5. Validation of simulations 91

(a) Primary image (b) Scatter image (c) SPR

(d) Primary image (e) Scatter image (f) SPR

Figure 4.12: Images recorded within the image receptor for Experiment I. The first row corresponds to images with anti-scatter grid present, and the second row illustrates images without the anti-scatter grid. The red square shown in (a) represents the location of the ROI used to calculate CIF and BF.

The ROI used in this work to calculate the SPR is shown in Figure 4.12(a) (red square). It is unknown, where Boone et al. measured the SPR. This could also lead to discrep- ancies, as SPR changes across the detector when the anti-scatter grid is not employed. Moreover, Boone et al. has not included the scatter originated in the anti-scatter grid, whereas in this work, scatter from all elements in the geometry (including anti-scatter grid) is simulated.

Table 4.8: Experiment I: SPR, CIF and BF values calculated from this

work and Boone et al.(2002) [61]. SEM associated to each calculation of this work is shown in brackets.

Boone et al.(2002) This work Difference Grid No Grid Grid No Grid Grid No Grid SPR — 0.690 0.0749 (0.2%) 0.689 (0.1%) — 0.1% CIF 1.54 1.57 (0.0%) 2.0%

92 Chapter 4. Monte Carlo simulations of X-ray mammography

Table 4.9: Experiment II: SPR, CIF and BF values calculated from this work and Boone et al.(2002) [61]. SEM associated to each calculation of this work is shown in brackets.

Boone et al.(2002) This work Difference Grid No Grid Grid No Grid Grid No Grid SPR — 0.660 0.135 (0.1%) 0.610 (0.0%) — 7.6% CIF 1.44 1.42 (0.0%) 1.4%

BF 1.60 1.57 (0.0%) 2.0%

Table 4.10: Experiment III: SPR, CIF and BF values calculated from this work and Boone et al.(2002) [61]. SEM associated to each calculation of this work is shown in brackets.

Boone et al.(2002) This work Difference Grid No Grid Grid No Grid Grid No Grid SPR — 0.970 0.120 (0.4%) 0.920 (0.3%) — 5.2% CIF 1.75 1.71 (0.2%) 2.3%

BF 1.95 1.93 (0.3%) 1.0%