Evaluation of a simplified optimizer for MR‐guided adaptive RT in case of pancreatic cancer

Abstract Purpose Magnetic resonance‐guided adaptive radiotherapy (MRgART) is considered a promising resource for pancreatic cancer, as it allows to online modify the dose distribution according to daily anatomy. This study aims to compare the dosimetric performance of a simplified optimizer implemented on a MR‐Linac treatment planning system (TPS) with those obtained using an advanced optimizer implemented on a conventional Linac. Methods Twenty patients affected by locally advanced pancreatic cancer (LAPC) were considered. Gross tumor volume (GTV) and surrounding organ at risks (OARs) were contoured on the average 4DCT scan. Planning target volume was generated from GTV by adding an isotropic 3 mm margin and excluding overlap areas with OARs. Treatment plans were generated by using the simple optimizer for the MR‐Linac in intensity‐modulated radiation therapy (IMRT) and the advanced optimizer for conventional Linac in IMRT and volumetric modulated arc therapy (VMAT) technique. Prescription dose was 40 Gy in five fractions. The dosimetric comparison was performed on target coverage, dosimetric indicators, and low dose diffusion. Results The simplified optimizer of MR‐Linac generated clinically acceptable plans in 80% and optimal plans in 55% of cases. The number of clinically acceptable plans obtained using the advanced optimizer of the conventional Linac with IMRT was the same of MR‐Linac, but the percentage of optimal plans was higher (65%). Using the VMAT technique, it is possible to obtain clinically acceptable plan in 95% and optimal plans in 90% of cases. The advanced optimizer combined with VMAT technique ensures higher target dose homogeneity and minor diffusion of low doses, but its actual optimization time is not suitable for MRgART. Conclusion Simplified optimization solutions implemented in the MR‐Linac TPS allows to elaborate in most of cases treatment plans dosimetrically comparable with those obtained by using an advanced optimizer. A superior treatment plan quality is possible using the VMAT technique that could represent a breakthrough for the MRgART if the modern advancements will lead to shorter optimization times.

possible using the VMAT technique that could represent a breakthrough for the MRgART if the modern advancements will lead to shorter optimization times.

K E Y W O R D S
MR-guided Radiotherapy, online adaptive radiotherapy, plan optimization, pancreatic cancer, IMRT

| INTRODUCTION
To date, pancreatic cancer represents one of the most aggressive tumors with a 5 years overall survival (OS) rate ranging from 5% to 20%, depending on stage at diagnosis. Surgery still represents the most valuable treatment option, although only 20% of patients appears to be candidate for resection at diagnosis. 1,2 The large majority of these patients presents indeed unresectable locally advanced tumors, whose clinical management is much more complex and characterized by very poor prognosis. Different clinical trials have demonstrated that hypofractionated radiotherapy (RT) combined with chemotherapy may improve OS for patients affected by LAPC, even if surrounding organs at risk (OARs) toxicity still remains a strong dose limiting factor in this setting. 3,4 The recent development of hybrid magnetic resonance-guided adaptive radiotherapy (MRgART), characterized by high soft-tissue contrast images, allows for online modification of the dose distribution taking into account the daily positions of the treatment volumes, thus allowing to safely deliver high doses to the target and minimizing the dose to OARs. 5,6 To date, two MR-Linac systems are available for clinical practice: MRIdian (ViewRay Inc, Mountain View, CA, USA) which couples a 0.35 T MRI scanner with a 6 MV Flattening Filter Free (FFF) Linac and Unity (Elekta, Stockholm, Sweden) which mounts a 1.5 T MRI scanner and a 7 MV FFF Linac. [7][8][9] Both systems are based on a transverse geometry system, where the magnetic field (B) force lines are transversely oriented with respect to the radiation beam axis.
The innovative solutions offered by the hybrid systems allow the delivery of online adapted treatment plans: to this end the speed of plan optimization becomes crucial, as the patient is waiting on couch during the entire adaptive procedure (that can last up to 50 min) and unnecessary delays could cause anatomical changes, jeopardizing plan integrity or exhausting patient's endurance in treatment position. 10 The MRIdian Linac treatment planning system (TPS) to date offers two solutions to optimize the treatment plans. The most common version used during adaptive procedures implements a simplified optimization algorithm based on a convex nonlinear programming model that aims to build a simple cost function with one global minimum. Having calculation speed as priority, this optimizer allows to put in the cost function only two parameters for the target volume (minimum and maximum dose) and one maximum dose constraint for each considered OAR. No dose-volume constraints can be put in this version of optimizer. 11 A new version was recently introduced, that allows to consider more than one constraint for OARs and more than two constraints for the target volume. Although it offers also the opportunity to put in both dose constraints and dose objectives, its application during the adaptive procedure is still quite limited, as a quick solution of the cost function is not always reachable.
The recent introduction of these fast and simplified plan optimizers leads to the necessity to test their ability to elaborate dose distributions comparable with those created by traditional advanced optimizers, which can consider many dose to volume constraints in their cost function.
In this study, the RT treatment plans generated using the simpli-    Table 1.
The treatment plans were generated on both systems using the intensity-modulated radiation therapy (IMRT) step and shoot CUSUMANO ET AL.

| 21
technique with 20 equidistant angular beams. For the traditional Linac, the treatment plans were also optimized using the volumetric modulated arc therapy (VMAT) technique with dual arc.
No VMAT solutions are currently available on MRI Linac: the high number of IMRT beams was therefore chosen to mimic a VMAT dose distribution.
All treatment plans were performed by the same planner in order to avoid the inter-operator variability bias. Prescribed dose was 40 Gy in 5 fractions to PTV for all the plans.
The dose distributions were calculated using the Analytical Anisotropic Algorithm for VMAT and IMRT traditional Linac plans and a Graphical Power Unit-accelerated Montecarlo algorithm for the MR-Linac plans. 12 An isotropic grid size of 2 mm was set for dose calculation in both systems, as recommended by AAPM TG101 for SBRT treatment plans. 13 The fluence map obtained during the optimization step on the MR-Linac TPS was calculated using a spatial resolution of 2 mm and considering the presence of the magnetic field.
The MR-Linac treatment plans were performed using the simpli- The mean time needed to optimize and calculate the different treatment plans was measured, excluding the time necessary to set the beam geometry and to create the supporting structures necessary to optimize the dose distribution, as they have been already created in a typical online adaptive scenario. The time necessary to obtain a first dose distribution was separated to that one required to reach the final solution.
All contours were double checked by two radiation oncologists on both planning systems, in order to reduce the uncertainties related to the different interpolation and smoothing algorithms adopted by the two TPSs.

2.B | Plan optimization strategy
The cases were classified in two different planning categories, considering the distance between the PTV outer margin and the closest OAR, in order to take into account the clinical heterogeneity observable in LAPC radiation treatment. This classification has been applied considering the distance between the PTV outer margin and the closest OAR and setting a threshold of 3 mm.
OARs constraints had to be respected in all the cases and treatment plan optimization was stopped only when one organ at risk maximum constraint was reached.

2.B.1 | Planning category 1
In the cases where the closest OAR had a distance to the PTV inferior to 3 mm, the primary objective was to obtain a CTV coverage higher or equal to 99.5 % with the 95% of the prescription dose (CTV V95% ≥ 99.5%), ensuring in the meanwhile the D1% of CTV inferior to 50 Gy. The secondary objective was to obtain the highest PTV coverage respecting all the OARs dose constraints.

2.B.2 | Planning category 2
In the cases where the minimum distance between the PTV margin and the OARs was larger than 3 mm, the objectives of the planning category 1 were considered as preliminary conditions to be achieved.
For these more anatomically advantageous cases the primary planning objective was to obtain a PTV coverage higher or equal to 98 % with the 95% of the prescription dose (PTV V95% ≥ 98%).
For both categories, very tight constraints in terms of target coverage were adopted, in order to ensure the achievement of a full optimization of the treatment plans.
All the plans were normalized setting the prescription dose at 50% of the PTV and allowing a D1% of PTV not higher than 50 Gy. Table 2 reports the dose constraints considered for the OARs for pancreatic MRgART, as proposed by Bouhoudi and colleagues. 14

2.C | Plan evaluation
All plans were firstly evaluated by two radiation oncologists to assess their clinical acceptability with the following criteria: 95% of the PTV had to be covered with at least 95% of the prescription dose and OARs constraints had to be successfully met. 15 In particular, considering the two planning categories defined, a treatment plan was indicated as "clinically acceptable" if the V95% of CTV was higher or equal to 99.5% and the V95% of PTV was comprised between 95% and 98%. A plan was labeled as "optimal" if two conditions were simultaneously reached: CTV V95% ≥ 99.5% and PTV V95% ≥ 98%. All the cases where CTV V95% was < 99.5% or PTV V95% was < 95% were considered as not clinically acceptable plans.  The overall dosimetric comparison of the dose distributions obtained from the two systems was based on target coverage, OARs sparing and low dose irradiated volumes.
The results were compared using the Wilcoxon signed-rank test and differences were considered statistically significant for P-value lower than 0.05. 16 Three dosimetric indicators were adopted to evaluate the quality of the dose distributions.
The conformity index (CI) was used to evaluate the conformity of the dose distribution on the PTV 15 : where V38Gy is the volume of the 95% of the prescription dose (38 Gy), while V(PTV) is the volume of the target structure.

| 23
The homogeneity index was calculated for CTV and PTV as defined by ICRU 83: where D2%, D98% and D50% represent the absolute doses covering the 2%, the 98%, and the 50% of the target structures, respectively.
The low-dose spread was evaluated considering the volumes in cubic centimeters (cc) of the 10 and 20 Gy doses and in terms of gradient index (GI) 17 : For nine cases analyzed the minimum distance between the PTV margin and the closest OAR was smaller than 3 mm. A schematic representation of the results obtained for the clinical evaluation of the treatment plans is reported in Fig. 1.
The simplified optimizer of the MR-Linac generated clinically acceptable plans in 80% of cases (16/20) and optimal plans in 55% of cases (11/20).
The number of clinically acceptable plans obtained using the advanced optimizer of the conventional Linac with IMRT technique was the same as that one obtained with the MR-Linac, but the percentage of optimal plans (65%) was higher.
Using the traditional Linac with the VMAT technique only in one case out 20 the treatment plan generated did not reached the clinical acceptability and in 90% of cases optimal plans were generated.
For patient 6 all the approaches failed in calculating a clinically acceptable plan, as the tumor and critical structures geometry resulted to be unfavorable for SBRT delivery.
The mean time required to obtain an initial dose distribution that solve the cost function was 1.6 ± 0.5 min for the MR-Linac, 1.8 ± 0.7 min for the IMRT with traditional Linac and 8.1 ± 1.8 min for the VMAT Linac.
The mean time to obtain the dose distribution that meets all the dose constraint was: 5.6 ± 1.7 min for the MR-Linac, 5.4 ± 2.1 min for the IMRT with traditional Linac and 13.6 ± 2.9 min for the VMAT Linac.  F I G . 1. Schematic representation of the results obtained in terms of clinical evaluation for the MR-Linac in IMRT modality (MR-Linac) and the traditional Linac in IMRT (IMRT) and VMAT (VMAT) modality. A plan was considered as "clinically acceptable" if the V95% of CTV ≥ 99.5% and V95% of PTV was comprised between 95% and 98%. A plan was labeled as "optimal" if CTV V95% ≥ 99.5% and PTV V95% ≥ 98%. All the other cases were considered as not clinically acceptable plans.
The volumes of V10 Gy and V20 Gy were significantly lower for VMAT respect the two IMRT plans (P < 0.001 and P = 0.002 for V10Gy and V20Gy comparing VMAT and MR-Linac; P = 0.001 and P < 0.001 between VMAT and IMRT with traditional Linac).
Lastly, no significant difference was observed among the different techniques for the OARs.

| DISCUSSION
Several studies have recently investigated the feasibility of generating clinically acceptable treatment plans for hybrid MRI guided treatment machines in different cancer sites, such as rectum, lung, and brain. 19,20 In particular, Ramey  It should be also observed that this study was carried out considering equal therapy volumes, but the high soft-tissue contrast

| CONCLUSIONS
Pancreatic cancer SBRT represents a challenge for RT delivery technologies for several reasons, such as the proximity between the tumor and radiosensitive OAR, and the significant inter-and intrafraction anatomical variability (e.g., different organ filling of hollow organs, breathing induced organ motion).
The MR-Linac system, thanks to the high soft-tissue contrast provided by the MRI and the possibility to adapt and gate the treatment by using real-time cine MRI, appears to be a good system to address these sources of variability.
This study demonstrates that the fast-simple optimizer implemented in the MR-Linac TPS allows to elaborate IMRT treatment plans with dosimetric performances comparable to those obtained by using a traditional Linac with an advanced optimizer and same delivery technique. A superior treatment plan quality is possible using the VMAT technique that could represent a breakthrough for the MR-guided Radiotherapy if the modern advancements will lead to shorter optimization times.
These results open new frontiers towards clinical approaches that aim to escalate the dose on target volumes while effectively sparing the surrounding healthy tissues.

CONFLI CT OF INTEREST
The authors have no relevant conflict of interest to declare.