Refining complex re‐irradiation dosimetry through feasibility benchmarking and analysis for informed treatment planning

Abstract Purpose/Objectives The purpose of this study is to dually evaluate the effectiveness of PlanIQ in predicting the viability and outcome of dosimetric planning in cases of complex re‐irradiation as well as generating an equivalent plan through Pinnacle integration. The study also postulates that a possible strength of PlanIQ lies in mitigating pre‐optimization uncertainties tied directly to dose overlap regions where re‐irradiation is necessary. Methods A retrospective patient selection (n = 20) included a diverse range of re‐irradiation cases to be planned using Pinnacle auto‐planning with PlanIQ integration. A consistent planning template was developed and applied across all cases. Direct plan comparisons of manual plans against feasibility‐produced plans were performed by physician(s) with dosimetry recording relevant proximal OAR and planning timeline data. Results and Discussion All re‐irradiation cases were successfully predicted to be achievable per PlanIQ analyses with three cases (3/20) necessitating 95% target coverage conditions, previously exhibited in the manually planned counterparts, and determined acceptable under institutional standards. At the same time, PlanIQ consistently produced plans of equal or greater quality to the previously manually planned re‐irradiation across all (20/20) trials (P = 0.05). Proximal OAR exhibited similar to slightly improved maximum point doses from feasibility‐based planning with the largest advantages gained found within the subset of cranial and spine overlap cases, where improvements upward of 10.9% were observed. Mean doses to proximal tissues were found to be a statistically significant (P < 0.05) 5.0% improvement across the entire study. Documented planning times were markedly less than or equal to the time contributed to manual planning across all cases. Conclusion Initial findings indicate that PlanIQ effectively provides the user clear feasibility feedback capable of facilitating decision‐making on whether re‐irradiation dose objectives and prescription dose coverage are possible at the onset of treatment planning thus eliminating possible trial and error associated with some manual planning. Introducing model‐based prediction tools into planning of complex re‐irradiation cases yielded positive outcomes on the final treatment plans.

The role of radiation within the cancer care continuum has become increasingly well understood in the 20 th and 21 st centuries, with studies suggesting a clinical indication for irradiation in over 50% of all diagnosed cases. 3,4 Furthermore, the advent and commercial use of treatment planning solutions, such as multi-criteria optimization (MCO), knowledge-based planning, AP, and PlanIQ, have enhanced practitioner's decision-making and reduced inter-and intra-planner variation. 5 Clinical researchers have also demonstrated powerful applications for PlanIQ and automated planning in cases of head and neck, hippocampal sparing, and stereotactic ablative body radiotherapy (SBRT), 6,7,8 . However, a gap in literature illustrating the true breadth of cases in which these treatment planning solutions are applicable has been identified. Consequently, this paper postulates the idea that a possible strength of PlanIQ rests in its ability to mitigate pre-optimization uncertainties tied directly to the dose overlap regions in cases of re-irradiation, presenting the planner with meaningful choices on how to improve plan optimization through the feasibility of the various constraints dosimetrists provide the software.
Purposely designed to replicate the actions taken by experienced planners, Pinnacle's AP has been shown to generate clinically acceptable treatment options by leveraging an iterative approach to optimization. 6,9 Additionally, AP allows users to develop templates, referred to as treatment techniques, for generalizable planning scenarios upon which planners may manually optimize and/or adjust at a later time. Manual adjustment, in this sense, is often exhibited in trial and error efforts to achieve a satisfactory outcome. An important limitation to note, as illustrated by Kumar 10 and Esposito,11 is that AP quality is highly dependent on the initial user request(s) and/or planner experience. In isolation, AP does not eliminate the uncertainty of whether a clinical objective is achievable or not. As a focal point within this study, the uncertainties of re-irradiation further exacerbate the limitations of automated planning functionality.
Representing the second core component within this research, PlanIQ offers a robust means of assessing patient anatomy, idealistic dose distributions, and clinical objectives prior to optimizing within a TPS. 12 Considered a highly valuable and validated tool for dosimetric planning, Plan IQ utilizes a priori estimation of each clinical goal, returning a feasibility score that planners can interpret, mitigating historical uncertainties, and optimization backtracking. 13,14 More specifically, PlanIQ returns users' optimization objectives as values along a feasibility dose-volume histogram (FDVH) that can be examined and/or manipulated to achieve ideal sparing of proximal tissues through Pinnacle optimization. 15 Recently, this has become a seamless process, as Pinnacle autoplan now offers direct export and import options for PlanIQ. Therefore, it is at the intersection of these two treatment planning solutions that we begin to develop a newfound sense of informed treatment planning for cases of re-irradiation.

| PURPOSE
The purpose of this study is to dually evaluate the effectiveness of PlanIQ in predicting the outcome of dosimetric planning in cases of complex re-irradiation as well as generating a treatment plan of equivalent quality through Pinnacle AP integration. Through the model-based system PlanIQ incorporates in its dose calculation methods, planners are presented with meaningful choices via PlanIQ Dose Analysis and Benchmarking from which this study attempts to integrate and refine the processes involved in the treatment planning of complex re-irradiation cases.
Historically, planning of re-irradiation cases has proven to be a time-consuming process for both physicians and planners alike. 16 Often times, trial and error plays a large role in determining dose objectives that are achievable within a planning system based solely on each individual planner's results for a given iteration of the optimization process, 17 whereas PlanIQ now looks to mitigate pre-optimization uncertainties in defining objectives and providing much needed clarity for dosimetrists through feasibility histograms. Specifically, in cases of re-irradiation where the dose-overlap regions directly affect the decision-making process of both physicians and treatment planners, dose feasibility benchmarking may provide the insight necessary to eliminate the trial and error of blind optimization all together. As a novel approach to examining and generating a reirradiation treatment plan, the integration of PlanIQ and AP allows for informed treatment planning that was previously inaccessible and/or based on trial and error. Not unlike earlier suggestions, the scope of this investigation is not to determine the finite best planning solution but rather to examine PlanIQ and its ability to enhance the dosimetrists planning knowledge through a system wherein guidance on achieving dose objectives is clearly relayed to the user as a metric of feasibility. 18 This study also aims to provide direct DVH comparisons and statistical analyses between the initial manually planned outcomes and the AP/PlanIQ outcomes. These comparisons are valuable because all of the plans evaluated within PlanIQ have been previously achieved through manual planning alone whereas this study now looks to refine and advance the process of planning re-irradiation through the use of AP and feasibility software. As previously noted by Fried et al., 13 complex treatment planning has an inherent degree of variability that rests largely with the user and their interpretation of how delineated target volumes and dose objectives interact. In the novel examination of re-irradiation cases, the integration of AP and PlanIQ mitigates the user-related variability with respect to previously delivered dose and the perceived influence of that dose with newly outlined treatment objectives. This perception leads to blind optimization in the absence of a feasibility planning solutions such as PlanIQ, which this study believes may exacerbate an already challenging planning scenario.

3.A | Patient selection & setup
Patient selection (n = 20) included a diverse range of re-irradiation cases from University Hospitals Seidman Cancer Center in Cleveland, Ohio. Re-irradiation cases, ranging from the brain to the lower extremities, are expected to provide an examination of the scope of applications in which feasibility benchmarking is applicable, rather than a snapshot of one specific anatomical site. All patients were simulated head first supine using a Siemens Somatom Sensation CT scanner (Siemens Medical Solutions, Malvem, PA) using 1-3 mm slice thickness. Immobilization techniques varied based on the desired site of treatment and patient reproducibility.

3.B | Target delineation
Pinnacle TPS (version 14.0 -16.2) and MIM (version 6.6) (MIM Software, Cleveland, OH) were used by the attending physician(s) to delineate both normal organs at risk (OAR) and approved target volumes. Variable target margins were used at the physician's discretion due to anatomical restrictions and proximity to sites of prior irradiation.
Dosimetry contouring was minimal and consisted primarily of segmenting OAR based on previously delivered doses which would then be exported to PlanIQ. These optimization target volumes, generated to account for direct overlap with previously irradiated critical structures, were used in two (2/20) cases to achieve a plan of equal quality with the manual plan. Given the prior treatment planning that had occurred, normal OAR were not modified and were evaluated as they were at the time of initial planning review. Artifacts produced by metallic implants, fillings, or fiducials were manually assigned a density matching that of the nearby tissues, bone, or adipose. The previously delivered doses were accounted for through an MIM dose accumulation which were then sent to Pinnacle as a contour set, providing a surrogate for the dose cloud which cannot be summed in PlanIQ at the time of this research. Proximal OAR overlapping the initial sites of radiation were segmented into contours that corresponded to the remaining dose they could receive without violating the chosen dose threshold. An oversimplified example being a region of spinal cord that previously received 10 Gy would be segmented out as "Spinal Cord MAX 35Gy," thus reflecting the remaining dose that could be delivered before exceeding the generally accepted tolerance dose [ Fig. 1]. 19 All overlapping OAR segments were entered into AP as a way of informing the planning system of the previously delivered dose and the strict constraints that would be necessary to accomplish the desired re-irradiation. Treatment planning objectives were entered into the AP window without modi-  20 In cases of re-irradiation, it is not uncommon to see critical structures, such as the spinal cord, overlapping both a previously irradiated volume and the newly delineated target. When this occurred, PlanIQ could return our requests as "Impossible" (Feasibility (F) < 0.0) or highly "Difficult" (F < 0.1). These requests needed to be met however, and therefore saw the "Compromise" function turned off within Pinnacle's AP. Some objectives were determined to be "Challenging" (F = 0.11-0.5) while other objectives, seemingly easier to achieve requests, were deemed "Probable." All

| RESULTS & DISCUSSION
A side-by-side comparison of isodose lines and dose-volume histograms (DVHs) was performed by the researching physicians to establish whether the PlanIQ-AP generated plan was acceptable.
Once reviewed and/or approved, all dose information was recorded for target volumes, critical OAR, dose overlap regions, and all contoured/segmented areas of interest.
Not only does PlanIQ provide physicians and dosimetrists with pre-optimization insight into generating and pursuing realistic treatment objectives in these cases of complex re-irradiation, but it can also provide feedback that has the potential to drive important decision-making surrounding planning prescriptions. Historically, blind optimization would be performed via trial and error until it was determined one prescription would or would not be achievable. This study found that the feasibility software provided sufficient information to inform whether one prescription would be achievable at the onset of planning, saving significant planning time, and allowing dosimetrists to focus on achieving outlined goals while simultaneously delivering acceptable levels of prescription dose coverage to delineated target volumes. It should be noted that a small margin can be added to proximal OAR and, under re-irradiation conditions specifically, subtracted from F I G . 2. The left-hand column represents the template described in methods and materials. Continuity in planning approach was necessary given the degree of variation between the retrospectively planned cases. The right-hand column demonstrates feasibility DVHs (FDVHs) and the respective F-Values for a given curve.
the PTV in order to attain a planning target structure. This allows AP to generate a steep dose gradient toward the structure whose dose objective must be observed over target coverage, such as a spinal cord approaching threshold dose. Furthermore, the six cases utilizing optimization target volumes for PlanIQ-AP benchmarking and optimization had final plans determined to be of equal quality when compared to the manual plan. Similar to the findings of, 21 optimization of OAR close to targets, or overlapping in the case of re-irradiation, presented planners with an opportunity to benefit greatly from manipulating FDVHs for these segmented OAR.
OAR within regions of direct dose overlap saw improvements in 17 (17/20) patients, with the remaining three patients exhibiting similar, or unremarkable, improvements compared to the manual plan. Average maximum point doses to critical OAR for plans utilizing PlanIQ-AP saw improvements ranging from as low as 60 cGy to as high as 5 Gy (brain) and an overall improvement of eight (8%) percent when compared to manual planning. These findings were found to be statistically significant (P < 0.06). Additional PlanIQ-AP improvements, observed in patient example [ Fig. 3], in max point dose and integral doses to proximal tissues and OAR can be found within specific dose overlap regions. Additional significant comparisons (P < 0.05) of normalized mean dose to proximal OAR were found to exhibit an overall 5% improvement through the use of Pla-nIQ-AP [ Fig. 4]. Utilizing PlanIQ's feasibility feedback into ideal Consequently, this paper recommends that future studies consider the planning and examination of cases across multiple treatment centers and/or facilities to validate the template developed and presented by University Hospitals Seidman Cancer Center.

| CONCLUSION
Emerging technologies continue to enhance practitioner's abilities to provide patient-specific treatment options, with the integration of PlanIQ and AP representing a milestone in the planning of complex re-irradiation cases. Based on the initial findings of this study, it is clear that PlanIQ-AP effectively provides the necessary feedback to facilitate physician decision-making and eliminate blind optimizations through trial and error. Additionally, findings also suggest significant reductions in integral mean doses to proximal OAR as well as maximum point doses within critical dose overlap regions. To our knowledge, novel planning solutions that can adequately contend with such a wide array of re-irradiation cases have not been fully realized and/or examined within the scientific, peer-reviewed, setting. To that extent, it is the introduction of model-based prediction tools into AP that has yielded positive outcomes in treatment plan generation and workflow efficiency. Rodney J. Ellis served as the senior author, senior investigator, physician reviewer, editor, project coordinator, and mentor.

CONFLI CT OF INTEREST
While University Hospitals Seidman Cancer Center uses Elekta linear accelerators, MOSAIQ record and verify EMR, Pinnacle treatment planning system, and MIM, we have no conflict of interest to disclose.