Personalized setting of plan parameters using feasibility dose volume histogram for auto‐planning in Pinnacle system

Abstract Purpose The personalized setting of plan parameters in the Auto‐Planning module of the Pinnacle treatment planning system (TPS) using the PlanIQ feasibility tool was evaluated for lung cancer conventional fractionated radiotherapy (CFRT). Materials and method We reviewed the records of ten patients with lung cancer who were treated with volumetric modulated arc therapy (VMAT). Three plans were designed for each patient: the clinically accepted manual plan (MP) and two automatic plans including one generated using the generic plan parameters in technique script (AP1) and the other generated using personalized plan parameters derived based on feasibility dose volume histogram (FDVH) in PlanIQ (AP2). The plans were assessed according to the dosimetric parameters, monitor units, and planning time. A plan quality metric (PQM) was defined according to the clinical requirements for plan assessment. Results AP2 achieved better lung sparing than AP1 and MP. The PQM value of AP2 (52.5 ± 14.3) was higher than those of AP1 (49.2 ± 16.2) and MP (44.8 ± 16.9) with P < 0.05. The monitor units of AP2 (585.9 ± 142.9 MU) was higher than that of AP1 (511.1 ± 136.5 MU) and lower than that of MP (632.8 ± 143.8 MU) with p < 0.05. The planning time of AP2 (33.2 ± 4.8 min) was slightly higher than that of AP1 (28.2 ± 4.0 min) and substantially lower than that of MP (72.9 ± 28.5 min) with P < 0.05. Conclusions The Auto‐Planning module of the Pinnacle system using personalized plan parameters suggested by the PlanIQ Feasibility tool provides superior quality for lung cancer plans, especially in terms of lung sparing. The time consumption of Auto‐Planning was slightly higher with the personalized parameters compared to that with the generic parameters, but significantly lower than that for the manual plan.


| INTRODUCTION
Along with software advancements to handle increasingly complex calculations, radiation treatment planning is rapidly becoming more and more automated to relieve the planner of tasks that can be readily handled by computers, such as autosegmentation and optimization. 1 After localization images of the target are obtained and exported to a treatment planning system (TPS), most of the steps involved in generating the anatomical structure and optimizing the treatment plan can be automated. During the optimization process, the use of automatic planning can reduce human variability while maintaining similar treatment plan quality.
The quality of a manual radiotherapy treatment plan strongly depends on the planner's experience and the planning time 2 4 Some studies imply that planning automation can also be useful for adaptive radiation therapy 5 and unbiased comparisons of treatment techniques. [6][7][8] The AP module has already been tested in developing plans for head and neck treatments, [9][10][11][12] hippocampal avoidance whole-brain treatment, 13 prostate treatments, 14 and liver tumor stereotactic body radiotherapy (SBRT). 15 The results of these studies show that this approach may improve the efficiency of the optimization process, eliminate the need for repeated trial-and-error during the manual planning process, and tend to improve the standardized plan quality.
In the AP module, it is still necessary to manually set the beam angles, prescription, and the initial optimization parameters in advance. Generally, the selected initial optimization parameters directly affect the final plan quality. For cases with the same clinical requirements, the AP module uses generic initial parameters in the technique script according to in-house planning experience. However, these generic initial parameters are not appropriate for some cases, necessitating manual modification via trial-and-error. PlanIQ Feasibility (Sun Nuclear Corp, Melbourne, FL) is a tool to estimate the best possible sparing dose of organs at risk (OARs) a priori before starting the plan optimization. 16 Assuming an ideal fall-off from the prescription dose at the target boundary, a feasibility dose volume histogram (FDVH) quantitatively determines the regions of a DVH that are impossible (red), difficult (orange), challenging (yellow), and probable (green) for each organ at risk (OAR). The FDVH allows the OAR planning goals to be personalized according to the patient geometry. In the latest version of Pinnacle v16.2, the PlanIQ Feasibility tool has been integrated into the AP module.
In this study, we evaluate an AP workflow in which the plan parameters are personalized using the PlanIQ Feasibility tool, and verify its effectiveness in developing VMAT plans for 10 lung cancer patients. For each case, three plans were developed: a clinically accepted manual plan (MP), an automatic plan generated using generic initial parameters (AP1), and another automatic plan generated using personalized OAR sparing goals according to the FDVH  Table 1. The prescribed dose was 60 Gy, delivered in 30 fractions.
The planning objectives for the PTV were that the relative volume that receives ≥ 100% of the prescribed dose > 95%, and the maximum point dose < 110% of the prescribed dose. The dose coverage and homogeneity of the PTV were assessed based on the dose distribution, the dose volume histogram (DVH), and the trade-off between the dose delivered to the PTV and OAR sparing. The planning objectives for the OARs were as follows: point dose, spinal cord < 40 Gy; point dose, spinal cord PRV < 45 Gy; volume of whole lung receiving more than 5 Gy (V5) not specified though lower doses are preferred, and that receiving more than 20 Gy (V20) < 28% and that receiving more than 30 Gy (V30) <20%; mean dose, whole lung (D mean ) <17 Gy; and volume of heart receiving more than 30 Gy (V30) < 40% and that receiving more than 40 Gy (V40) < 30%. The constraints specified in our department were T A B L E 1 Tumor staging.

Tumor staging
Number of patients mainly based on the quantitative analysis of normal tissue effects in the clinic (QUANTEC) guidelines, [19][20][21] but were more stringent.

2.B | Planning process
During the planning process, the arrangement of beams, dose prescription for the PTV, and initial optimization parameters of each OAR were set by loading a predefined technique script. The work- In the Feasibility tool, the adjusted optimization goals for the lungs were between difficult (orange) and challenging (yellow), and those for the heart and spinal cord were between challenging (yellow) and probable (green). In the AP module, the OAR priority parameters were set as follows (lung: high or medium, compromise; heart: low, compromise; spinal cord: high, no compromise).

2.C | Study endpoints
The three plans for each patient were compared in terms of the dosimetric parameters: PQM value, total monitor units, and planning time.
The dosimetric parameters are as follows: (a) The homogeneity index (HI) of the PTV was defined as follows: where D2%, D50%, and D98% are the minimum doses delivered to 2%, 50%, and 98% of the PTV, respectively. 22  of the PTV is defined as follow: where V PTV is the volume of the PTV, TV PTV is the portion of the V PTV that is within the prescribed isodose line, and the TV is the treated volume of the prescribed isodose line. 23  As shown in Table 2, a PQM scoring procedure with 10 related submetrics was defined for a treatment plan. Each metric is calcu- | 121

2.D | Statistical analysis
Wilcoxon signed rank tests were carried out to compare the MP, AP1, and AP2 for all obtained dosimetric data. The statistical analyses were performed in SPSS v17 (IBM Corp), with significance set at P < 0.05. Figure 2 shows an example of the distribution of the three plans (MP, AP1, and AP2). All three plans satisfied the clinical requirements for all OARs. Table 3 shows the average differences between MP, AP1, and AP2 among all metrics evaluated. There were statistically significant differences for the following parameters: PTV CI, lungs V20 Gy (%), V30 Gy (%), and D mean (Gy); and heart V30 Gy (%), Cord D max (Gy), and Cord PRV D max (Gy). For dose deposition in the lungs, the V20 Gy (%) and V30 Gy (%) of AP2 were lower than those of MP and AP1 (P < 0.05), and the D mean (Gy) of AP2 was lower than that of AP1 (P < 0.05). For dose deposition in the other OARs, the heart V30 Gy (%), cord D max (Gy), and cord PRV D max (Gy) of AP1 and AP2 were lower than those of MP (P < 0.05). Figure 3 shows the relationships of the metrics between AP1 and AP2.

| DISCUSSION
In this study, we evaluated a technique to personalize plan parameters for VMAT AP in treatment of patients with lung cancer. Three plans (MP, AP1, and AP2) were generated for each case. In AP2, the PlanIQ Feasibility tool was used to generate the personalized plan parameters.
The statistical results in Table 3 show that APs achieved better OAR sparing than the MPs. Furthermore, the dose deposition in the lungs was improved by using personalized parameters; while there was no statistically significant difference between MP and AP1, it appeared that the average lung dose in AP1 was slightly worse than that in MP, even though the priorities of the lung parameters in AP1 were set high. This could be attributed to the use of inappropriate initial settings in AP1 because, although MP and AP1 used the same initial plan parameters, the parameters in MP can be adjusted as The improvement in plan quality can be directly attributed to the higher degree of personalization of plan parameters.
The relationships of the metrics in Fig. 3 showed that AP2 provided superior protection of the lungs. Although there was no statistically significant difference in the CI and HI of the targets, it still can be observed from Fig. 3 that AP2 had inferior results on CI and HI compared to AP1. This could be attributed to the trade-off of OAR sparing for AP2. Thus, it is necessary to use a quantitative scoring procedure to assess the overall quality of a treatment plan.
The plan quality scoring procedure was defined based on concepts of PQM proposed by Benjamin 17 and the plan quality score, S D , proposed by Bohsung. 18 The evaluated metrics were the standard clinical OAR constraints for lung cancer treatment, and the score assignments of different metrics were determined by both physicians and planners. It should be noted that differences in PQM values are only meaningful between different plans for one patient. The features of lung cancer plans approved for clinical treatment showed that the HI and CI of the PTV can be properly sacrificed for OARs sparing, especially for decreasing the dose deposition in the lungs. 25 Thus, the score proportion of the OARs was higher than that of the PTV. However, the scoring procedure may differ between different tumor sites. For example, the radiotherapy plans for nasopharynx cancer (NPC) require better dose coverage and homogeneity of PTV than lung cancer plans, the score proportion of PTV should be increased. According to the plan quality scoring procedure, AP2 provided higher quality than either MP or AP1.
The MP generally has more monitor units than AP1 and AP2, Therefore, a priori estimated objectives should be chosen from difficult to probable regions based on the priorities of the OARs.
In AP2, the personalized parameters and priorities for the target and OARs were selected by trial-and-error. However, this might be a limitation because different initial settings lead to variable plan quality. Although we made an effort to find an appropriate initial setting, this selection may still result in suboptimal plan quality. To address this problem, we can generate several sets of initial parameters within a reasonable scope, and assign different priorities to target coverage and OAR sparing. Then, we can find a "sweet spot" according to these different initial settings. In this way, it is possible and beneficial to further improve the plan quality.
The results of this study indicate that CFRT plans using the AP2 approach are of superior quality to those using AP1 approach or generated manually. In future study, we will continue to test the performance of the AP2 method in SBRT plans. We expect that, by selecting appropriate personalized initial parameters, the AP2 method can achieve better plan quality than the conventional AP1 method. As the automation of planning process continues to improve, it is anticipated that the automatically generated plans will T A B L E 3 Averaged differences and 95% confidence interval between each two plans