Quantitative evaluation of the impact of heterogeneity correction on left breast cancer radiotherapy performed with respiratory gating

Corresponding author: Abdulhamid Chaikh; Department of Radiation Oncology and Medical Physics, University Hospital of Grenoble, France. Cite this article as: Fdhila MT, Gabelle-Flandin I, Balosso J, Chaikh A. Quantitative evaluation of the impact of heterogeneity correction on left breast cancer radiotherapy performed with respiratory gating. Int J Cancer Ther Oncol 2016; 4(1):417. DOI: 10.14319/ijcto.41.7


Introduction
Recent advances in radiotherapy have introduced several irradiation techniques for breast cancer. Several studies have demonstrated that respiratory gating improves the accuracy of dose delivery to the target volume in breast cancer radiotherapy by reducing doses to organs at risk (OAR), i.e the heart and lungs. 1,2 During the treatment, the dose received by the target volume (TV) should be close to the prescribed dose. The former is currently predicted by the treatment planning software based on dose calculation models. Certain organs i.e. lung and heart have tolerance dose limits that should be respected and an accurate dose calculation is necessary for this. Presently, there are various algorithms and methods to take account of the heterogeneity correction of tissues. Numerous reviews in the literature have categorized the heterogeneity correction methods according to various criteria. Recently, Chaikh et

Clinical cases and treatment planning
This study is based on ten patients with left breast cancer. These cases were chosen to cover the full range of the different types of radiotherapy, namely: breast alone, breast with boost, breast with internal mammary chain (IMC) or with IMC and the supra-clavicular area. A computed tomography (CT-scan) was done using the respiratory gating technique. Then the images were loaded into the TPS.
The virtual simulation for each patient was generated by a digitally reconstructed radiograph (DRR) and beam's eye view (BEV) information. In this way a treatment field was superimposed on the DRR to assess target localization. The target volumes including the security margins and the organs at risk were delineated by the radiation oncologist. The treatment was performed with four tangential beams according to the technique currently used in our department: 2 tangential beams of 6 Megavoltage (MV) and 2 tangential beams of 18 MV. The combination of four beams of two energies with different beam ponderation and field size, with or without a wedge, has been introduced to optimize the dose homogeneities more accurately than with just a pair of tangential beams with wedge filters. The conventional radiation course varied from 46.91 to 50.6 Gy in 23 -25 daily fractions of 1.87 -2.2 Gy on the whole breast followed by the addition of a boost to the tumor bed. The boost consisted of doses varying from 11 to 15 Gy in 5 -23 daily fractions of 0.5 to 2.5 Gy.
For each patient, two treatment plans were generated: in plan 1 the dose was calculated using the PBC which was taken as the reference dose, in plan 2 the dose was calculated using the MB method in combination with the PBC algorithm. The PBC algorithm is a convolution algorithm based on pencil beam kernels. First, the dose distribution in a homogeneous water equivalent medium is calculated. Next, the tissue densities are taken into account by multiplying the dose distribution by a correction factor. In the MB method, the inhomogeneity correction factor is calculated along the central axis.

Monitor units (MU)
A calculation of the average and standard deviation was performed for each field.

Isodose curves
The 95% and 100% isodose curves encompassing the TV were compared.

Dose volume histogram (DVH)
For each TV the minimum dose (Dmin), mean dose (Dmean), maximum dose (Dmax) and the calculated dose to 95% of the target volume (D95%) were compared using cumulative cDVH. For lung and heart the Dmin, Dmean and Dmax were compared. The dose homogeneity inside the TV, lung and heart were compared using a S-index associated with the differential dDVH 9 : where, D(j) is the relative dose in the lesion voxel j, Dmean is the average relative dose in the lesion and TV is the target volume in elementary voxels.
For each patient, to compare plan 2 with plan 1, the calculated doses in plan 1 were taken as the reference values. The difference in percentage was calculated as:

Quality indices
The coverage index (CI) and conformity index for the target volume (CITV) proposed by the Radiation Therapy Oncology Group (RTOG) were compared using the following equations 10,11 : where, Imin is the minimal isodose surrounding the target and RI is the reference isodose.

CITV =
(4) where, PD is the prescribed dose and TV is the target volume.

Variation of lung density using a gating technique
The CT images using either free breathing or the gating technique were exported to TPS, then the average hounsfield unit (HU) value was calculated for lung. Next, The HU was converted to relative electronic density (ρw) using the following equation 12 : where, ρw is the density relative to water measured using CT images. Figure 1 shows an example of a HU calculation using the TPS for a volume of interest located in the lung.

Gamma index
The impact of heterogeneity correction was analyzed using the two dimensional (2D) gamma index (γ) by comparing dose distributions from plan 1 and plan 2. 13, 14, 15 Our goal was to determine the volume ratio receiving the same irradiation, in terms of γ which combines two criteria including the dose difference (ΔDose) in percentage and the distance-to -agreement (DTA) in millimeters. An ellipse is used to determine the acceptable region, and γ ≤ 1 represents fulfillment of the criteria. The DICOM images for each patient were exported from TPS to RIT-113 ® (Dosimetry System Version 5.2, Radiological Imaging Technology, Inc., CO) including the dose distributions. The dimensions were 20 × 20 cm² with a resolution of 0.39 mm. The results were displayed using γ maps and dose pixel histograms (DPH). The γ maps show the pixels with γ values greater than unity that were outside of tolerance range. The pixels with γ > 1 show either an under-or overestimated dosage associated with plan 2 compared to plan 1. In order to discriminate between an over-and an under-estimated dose, a color-code was attributed to the ΔDose. The DPH indicates the fraction of pixels with a specific value of ΔDose. Using the γ criteria (3%, 3 mm) we considered that the plan 1 and plan 2 were similar if 95% of pixels had γ ≤ 1.

Statistical analysis
A Wilcoxon signed rank test was used to calculate the p-value. A bilateral statistical test was realized with an error α = 5%, corresponding to a 95% confidence interval. The Δdose is considered significant if p < 0.05. Data are presented as average ± standard deviation (SD). The statistical correlation between calculated doses from plan 1 and plan 2 was evaluated using Spearmen's correlation coefficient. 16 The correlation between density difference (Δdensity) and Δdose was also evaluated.

Monitor units
The MB reduced the MU by on average 1.12 ± 5.33%. The Wilcoxon test showed that there was no significant difference between plan 1 and plan 2, p = 0.27 and the data showed a strong correlation with r = 0.99.

Isodose curves
In the transversal plans, we observed that both the 95% and 100% isodose curves encompassed the same fraction of the TV with the two techniques. The 40% isodose curve is closer to the heart and lung with MB than with PBC. Figure 2 shows isodose curves in color shading showing the dose distribution from plan 1 with PBC and plan 2 with MB.    Figure 3 shows the cDVH from plan 1 and plan 2. The Tables 1 and 2 show dosimetric and statistical results obtained for TV, lung and heart. In Table 1, it can be seen that the Wilcoxon test indicated a very significant difference between calculated doses with PBC and MB.

cDVH
Spearman's test indicated a strong correlation for Dmin, Dmean and D95%. In Table 2, it can be seen that MB calculated higher doses for lung and heart with very significant differences, except for the lung's Dmin. The Spearman's test indicated a strong correlation except for the lung's Dmax.

dDVH
The dose distribution inside the TV, lung and heart using MB was slightly more heterogeneous than with PBC. The average S-indexes for the target were 10.52 ± 10.11 and 10.37 ± 10.04 for MB and PBC, respectively. Wilcoxon and Spearman's tests showed no-significant differences but strong correlations with p = 0.46 and r =0.99. For lung, the average S-indexes were 31.3 ± 4.19 and 29.28 ± 4.04 for MB and PBC, respectively. For heart, the average S-indexes were 1.89 ± 1.99 and 1.81 ± 2.01 respectively. Wilcoxon and Spearmen's tests showed a significant difference and strong correlation with p < 0.005 and r = 0.99.

Quality indices
The average CIs were 1.05 ± 0.58 and 0.98 ± 0.56 for MB and PBC, respectively, with p = 0.03 and r = 0.98. The average CITVs were 0.67 ± 0.27 and 0.63 ± 0.29 for MB and PBC, respectively, with p = 0.20 and r = 0.81. Figure  4 shows the indices values using a boxplot representation. It can be seen that both plan 1 and 2 have the same median values for CI and close values for CITV. Figure 4: Boxplot for quality indices indicating the minimum, median and maximum values, as well as the 25 th and 75 th percentiles.

Variation of lung density using a gating technique
The HU calculated for lung was -727.25 ± 42 and -835.25 ± 23.68 using free breathing and gating techniques, respectively. This leads to a decrease in lung density with the gating technique from 0.27 ± 0.04 to 0.16 ± 0.02. The Wilcoxon test showed a very significant decrease of density with p < 0.001. However, the data for density are highly variable from one patient to another, which explains why no correlation appears between the density and the associated breathing modality, free breathing or gating, r = 0.36. This is due to basal tissue density and a morphologic effect resulting in the gating technique being very different for each patient. Lung inflation during an inspiration, with gating increases the lung volume, which will consequently increase the distance between the TV and the heart. However, there was a poor correlation between density differences (Δdensity) and Δdose, with r values ranging from r = 0.52 to 0.62. Figure 5 shows the correlation between density and dose for Dmin, Dmean and Dmax for the TV for all patients.

2D gamma
The average maximum and mean γ values were 2.82 ± 0.65 and 0.23 ± 0.11, respectively. The tolerance limit of 95% of pixels having γ ≤ 1 is not respected at all. The average γ passing rate was 92.5 % ± 4.98 %. Figure 6 shows an example of a 2D γ map in the axial plane. The red and blue coloring indicate when the γ value > 1 and identify overestimated dose (DMB > DPBC) or underestimated dose (DMB < DPBC), respectively. Figure 7 shows the DPH obtained from plan 1 and plan 2. It can be seen that the tolerance limit of 95% of pixels having γ ≤ 1 can be satisfied with a ΔDose = 7% and 3 mm . The correlation test showed that there was a weak correlation between γ passing rates and dose difference.

Discussion
Using the respiratory gating technique, we observed that the lung density was decreased by 38.7 ± 9.82%. The range effect of lateral electrons is expected to be larger if the lung density is lower. The MB method integrated with PBC does not model the lateral electron transport. This algorithm will overestimate the coverage of the TV and underestimate the low dose volumes in or near to OAR. In this study, the MU for the set of treatments was substantially shorter with MB than with PBC, p > 0.05. The cDVH and dDVH analysis for the TV showed that MB overestimated the dose for the target volume with a significant difference, p < 0.05 (as shown in Table 1). Identifying the TV locations, we observed that the significant differences, up to 12%, were located in the internal mammary chain areas.
This study confirms, using the MB algorithm, that dosimetric parameters, derived from cDVH to lung and heart, will be increased (as shown in Table 2). The 2D γ analysis showed that the difference between plan 1 and plan 2 could reach up to ± 10%. Therefore, the tolerance level of 95% pixels with γ≤ 1 was not respected, showing a higher impact when using the density correction. Since respiratory gating is intended to reduce the dose to the heart and lungs, with the aim of reducing the risk of cardiac mortality and lung toxicity, one should take account of the alterations in dosimetry introduced by this technique, to retain its advantages. Thus, any change from PBC to newer algorithms, i.e. CCC, AAA or Acuros XB, and even more so in the context of respiratory gating, should be very carefully undertaken. 17-24

Conclusion
The most recent algorithms approximate lateral electron transport accurately, either in breast tissue with a density close to water, or lung with lower density. When using the gating technique for breast cancer treatment the lung density will be significantly decreased from 0.27 to 0.16, p < 0.001. The MB method for 1D heterogeneity correction overestimates the dose to the heart and lung. The physicist should be aware of the differences between the different dose calculation algorithms and should be careful when changing from older algorithms such as PBC to newer algorithms, in order to optimally protect the heart and lung.