On the reduction of aperture complexity in kidney SABR

Abstract Background Stereotactic ablative body radiotherapy (SABR) of primary kidney cancers is confounded by motion. There is a risk of interplay effect if the dose is delivered using volumetric modulated arc therapy (VMAT) and flattening filter‐free (FFF) dose rates due to target and linac motion. This study aims to provide an efficient way to generate plans with minimal aperture complexity. Methods In this retrospective study, 62 patients who received kidney SABR were reviewed. For each patient, two plans were created using internal target volume based motion management, on the average intensity projection of a four‐dimensional CT. In the first plan, optimization was performed using a knowledge‐based planning model based on delivered clinical plans in our institution. In the second plan, the optimization was repeated, with a maximum monitor unit (MU) objective applied in the optimization. Dose‐volume, conformity, and complexity metric (with the field edge metric and the modulation complexity score) were compared between the two plans. Results are shown in terms of median (first quartile — third quartile). Results Similar dosimetry was obtained with and without the utilization of an objective on the MU. However, complexity was reduced by using the objective on the MUs (modulation complexity score = 0.55 (0.50–0.61) / 0.33 (0.29–0.36), P‐value < 10−10, with/without the MU objective). Reduction of complexity was driven by a larger aperture area (area aperture variability = 0.68 (0.64–0.73) / 0.42 (0.37–0.45), P‐value < 10−10, with/without the MU objective). Using the objective on the MUs resulted in a more spherical dose distribution (sphericity 50% isodose = 0.73 (0.69–0.75) / 0.64 (0.60–0.68), P‐value < 10−8, with/without the MU objective) reducing dose to organs at risk given respiratory motion. Conclusions Aperture complexity is reduced in kidney SABR by using an objective on the MU delivery with VMAT and FFF dose rate.


| INTRODUCTION
Stereotactic ablative body radiotherapy (SABR) is a novel treatment to treat patients with renal cell carcinoma (RCC) for whom surgery is not an option. It results in excellent local control and low toxicity rates in primary RCC. 1,2 This technique is noninvasive and delivered in an outpatient procedure. Moreover, kidney SABR treatment is not limited to tumor size or kidney position as is the case with other alternatives to surgery such as radiofrequency ablation or cryoablation. 3 As the tumor moves during treatment with respiratory motion, the interplay between the moving multileaf collimator (MLC) and tumor motion may result in discrepancies between planned and delivered doses. Several studies have demonstrated limited impact of the interplay effect on conventionally fractionated treatments, as any discrepancies are averaged out during the course of treatment. [4][5][6][7][8][9][10] As opposed to conventionally fractionated treatments, interplay effect may impact SABR treatments because the number of fractions is typically 1-5. 10,11 Strategies to minimize interplay effect include reducing the aperture complexity, 11,12 increasing the number of beams and fractions, 10,12 reducing dose rate, 12,13 decreasing tumor amplitude 13,14 and breathing cycle [10][11][12]15 and treating with higher dose per fraction. 10,12,16 Contemporary kidney SABR treatments are often treated with volumetric modulated arc therapy (VMAT), often due to the challenging geometric relationship between the target and adjacent organs at risk (OAR). Furthermore, flattening filter-free (FFF) delivery is highly attractive to reduce the treatment time via increased dose rates. Due to substantial reductions in beam-on time with high dose rates, and the use of ultra-hypofractionated regimens, the interplay effect may impact the fidelity of the planned treatment dose. Furthermore, reduction of aperture complexity and reduction of required monitor units (MUs) are of interest to reduce the treatment delivery time, which may be particularly important in the context of respiratory gating or breath hold treatment delivery.
We have recently shown that reduction in aperture complexity can be achieved in lung SABR by using an optimization objective on the total number of MUs, referred as the "MU objective" in the Eclipse treatment planning system. 17 The MU objective has been available in the Eclipse treatment planning system for RapidArc optimization since version 8.5. Previous work established use of the MU objective results in reduction in the total MUs, and therefore in the beam-on time, while preserving adequate dosimetry by using the MU objective in prostate, 18,19 head and neck, 18,20 gynecological, 18 and lung SABR. 21 However, the impact on the aperture complexity by using the MU objective was not addressed in these previous studies.
The purpose of this study is to determine the dosimetric impact of reducing aperture complexity via inclusion of a penalty on total MUs in the optimization on kidney SABR VMAT treatment plans. It is hypothesized that substantial reductions in aperture complexity can be achieved with minimal impact on dosimetric quality in kidney SABR.

| MATERIALS AND METHODS
We included consecutive 62 patients with primary RCC treated with SABR between 2012 and 2018 at our institution. Fractionation was 26 Gy in a single fraction for lesion size smaller or equal to 4 cm and 42 Gy in three fractions for lesion size larger than 4 cm. 22 Out of the 62 patients, 23 patients received 26 Gy in a single fraction and 38 patients had 42 Gy in three fractions. One patient was treated at 18 Gy in one fraction, but was replanned in this study with 26 Gy in one fraction. These patients were treated with 3D conformal radiation therapy, intensity-modulated radiation therapy or VMAT. Ethics approval for this study was provided by Peter MacCallum Cancer Centre.
Each patient was simulated using a four-dimensional CT scan (4DCT). The tumor was segmented on all respiratory phases. Gross tumor volume (GTV) was accumulated on the average intensity projection (AIP) of the 4DCT to generate an internal target volume (ITV). A planning target volume (PTV) was created using an isotropic 5 mm expansion of the ITV. The AIP of the 4DCT was used for contouring, planning, optimizing, and calculating the dose distribution.
OARs were segmented on the AIP depending on the extent of respiratory motion. OAR contours in some cases overlap with the ITV contour in cases where the OAR is proximal to the tumor.
All patients were replanned for the purpose of this study. Plans were generated by using the Eclipse treatment planning system (Var- between 0 and 100. In addition to the MU objective, the aperture shape controller was set to "Very High", the convergence mode to "On", and the multiresolution (MR) level at restart to "MR3" for all plans.
Optimization and calculation was a two-step process to determine the upper value of the MU objective. The original MU was first obtained by optimizing and calculating the dose without the MU objective in one single process without user interaction. The plan was then normalized so that 100% of the prescription dose covers 95% of the target volume and is referred as "NMUO" plan in this study.
The NMUO plan was copied and an objective on the MU was added. The upper value of the MU objective was set to 50% of the original total MUs with a strength of 70, and the plan was then optimized from scratch through one process without user interaction.
The plan was also normalized so that 95% of the target volume is covered with 100% of the prescription, referred as "MUO" plan.
Any plan for which a dose constraint was not respected was replanned by adjusting the objectives to respect dose limits. These replans are referred as "modified KBP" as opposed with "original KBP" plan. These were all for cases in which an OAR was close to or overlapping with the target, and were identified as those with challenging plan geometry. If a structure was overlapping with the PTV or the ITV, the PTV and ITV structures were cropped to generate an optimization structure. In two patients, the two partial arcs of 210 o had to be modified to two full rotation arcs of 358°. The two calculation steps method was repeated by modifying the upper dose limit to the organ up to the point where all dose limits were respected. Where possible while meeting OAR constraints, the plans were normalized to 95% of the target volume was covered by the prescription dose. Where this was not possible, loss to target coverage was accepted to ensure OAR constraints were respected.
Dose metrics for the target and OARs, shown in Table 2, were evaluated. Dose limits were based on QUANTEC recommendations. [23][24][25] Plan generation, optimization, calculation, and metrics extraction were done by using the Eclipse Scripting Application Programming Interface (ESAPI). Plan conformity was determined using the RTOG conformity index, defined as the reference isodose volume divided by the target volume. 26 The 95% isodose was used as reference isodose to calculate the conformity index (CI95). A value of CI95 = 1 indicates ideal conformation. The target was partially irradiated if CI95 < 1 while the irradiated volume was greater than the target volume if CI95 > 1. 27 Moreover, acceptable CI95 values were defined as values smaller than 1.2 while minor deviations were defined for values greater than 1.2 but smaller than 1.5. 28 Conformity of the low dose region was assessed with the CI50, or equivalently the R50, defined as the 50% isodose volume divided by the target volume. CI50 conformity deviation was assessed through the ALARA principle with a planning goal of CI50 < 5 for all PTV volumes.
Due to the proximity of bowel structures to the target, and their variation in position between treatment planning and each treatment session, it is desirable to minimize higher isodose lines extending between bowel loops which may arise as a consequence of using the bowel structures for optimization. This was assessed by calculating the sphericity of the isodoses lines. The 100% and the 50% isodose lines were converted to contours and exported. Pyradiomics v3.0 29 was used to calculate the sphericity of the contour. The resulting value ranged between 0 and 1, where 1 indicated a perfect sphere.
Robustness of the plans were measured by calculating the edge metric (EM) and the modulation complexity score (MCS). EM was calculated according with C 1 = 0 and C 2 = 1. [30][31][32] In this representation, EM reports the y-leaf sides normalized by the area aperture weighted per control point. Plan complexity decreases as EM decreases to 0. MCS was interpreted according to McNiven et al. 33 and used by others. 31,34 MCS is a score based on adjacent leaf sides,

3.A | Dosimetry
An example of a typical dose distribution in the axial plane is shown in Fig. 1   Dose metrics for the whole multifraction cohort are shown in Near maximum dose to the small bowel was close to the dose limit in modified KBP plans with 60% (9/15) of plans having coverage reduced to meet this limit (small bowel D0.03cc = 28.8

3.B | Conformity Index and sphericity
CI95 is shown in Fig. 3 Isodoses were more spherical when using the MU objective.
An example of the 100% and 50% isodoses for the patient with the worst isodose 50% sphericity in NMUO plan is shown in    Table 3. Multifractionation subset plans were less complex than the single fraction subset plans. The type of fractionation was based on lesion size with the PTV larger in the multifractionation subset. This leads to a larger aperture area to cover the target and to a smaller EM and larger MCS. Using the MU objective decreased aperture complexity (relative difference of the medians with respect to NMUO plan was −61% and 64% in EM and MCS, respectively, both P-value < 10 −8 ). The two regimes are shown in Fig. 5(b). Even if some modulation is reduced in MUO plan, they are still more complex than their DCAT counterpart (relative difference of the medians in complexity of CF with respect to MUO plan was −61% and 50% in EM and MCS, respectively, both P-value < 10 −10 ).
The impact of using an objective to the MU is explained by AAVw and LSVw. These two quantities are shown in Fig. 5(b) while the median and interquartile range of the distribution and statistical significance are detailed in Table 3. Reduction in complexity was driven by a larger area aperture with the MU objective while leaf traveling changed minimally. This was expected as the parameter controlling the leaf travel, the so-called aperture shape control, was the same in both subsets. Using an objective on the MU reduces the degree of freedom available to the optimizer as the MUs were constrained. This loss was compensated by a larger aperture area. Therefore complexity was reduced.

| DISCUSSION
Results suggest that aperture complexity is minimized in kidney SABR treatment by using VMAT with FFF dose rate and an objective on the MU. Using an objective on the upper value MU forced the optimizer into using a larger aperture area. Since leaf travel was kept constant, aperture area increase resulted to a reduction in plan complexity. Importantly, there was minimal dosimetric compromise when  The main advantage of the objective on the MU is the beam-on time reduction that follows from the MU reduction (50% reduction in this study). Value of the upper limit on the MU objective and its associated strength were not optimized in this study, but were derived from previously investigated values in lung SABR. 35 The level of complexity achieved with VMAT and DCAT fields were similar in this previous study. This was not the case in this work as VMAT plans with the MU objective in use were still more complex than DCAT fields. This difference between the two studies may be due to the parameters used in the application of the MU objective

| CONCLUSION
In conclusion, kidney SABR plan quality is optimal by using an objec- where the sum runs over all the control points cp within the beam,

APPENDIX C SPHERICITY
The sphericity was determined with the Pyradiomics v3.0 module.
This quantity is defined as where V is the volume of the structure and A is the surface area of the structure. In these terms, the sphericity is a dimensionless quantity reporting value between 0 and 1, where 1 represents a perfect sphere.
GAUDREAULT ET AL.