Assessment of PlanIQ Feasibility DVH for head and neck treatment planning

Abstract Introduction Designing a radiation plan that optimally delivers both target coverage and normal tissue sparing is challenging. There are limited tools to determine what is dosimetrically achievable and frequently the experience of the planner/physician is relied upon to make these determinations. PlanIQ software provides a tool that uses target and organ at risk (OAR) geometry to indicate the difficulty of achieving different points for organ dose–volume histograms (DVH). We hypothesized that PlanIQ Feasibility DVH may aid planners in reducing dose to OARs. Methods and materials Clinically delivered head and neck treatments (clinical plan) were re‐planned (re‐plan) putting high emphasis on maximally sparing the contralateral parotid gland, contralateral submandibular gland, and larynx while maintaining routine clinical dosimetric objectives. The planner was blinded to the results of the clinically delivered plan as well as the Feasibility DVHs from PlanIQ. The re‐plan treatments were designed using 3‐arc VMAT in Raystation (RaySearch Laboratories, Sweden). The planner was then given the results from the PlanIQ Feasibility DVH analysis and developed an additional plan incorporating this information using 4‐arc VMAT (IQ plan). The DVHs across the three treatment plans were compared with what was deemed “impossible” by PlanIQ's Feasibility DVH (Impossible DVH). The impossible DVH (red) is defined as the DVH generated using the minimal dose that any voxel outside the targets must receive given 100% target coverage. Results The re‐plans performed blinded to PlanIQ Feasibilty DVH achieved superior sparing of aforementioned OARs compared to the clinically delivered plans and resulted in discrepancies from the impossible DVHs by an average of 200–700 cGy. Using the PlanIQ Feasibility DVH led to additional OAR sparing compared to both the re‐plans and clinical plans and reduced the discrepancies from the impossible DVHs to an average of approximately 100 cGy. The dose reduction from clinical to re‐plan and re‐plan to IQ plan were significantly different even when taking into account multiple hypothesis testing for both the contralateral parotid and the larynx (P < 0.004 for all comparisons). No significant differences were observed between the three plans for the contralateral parotid when considering multiple hypothesis testing. Conclusions Clinical treatment plans and blinded re‐plans were found to suboptimally spare OARs. PlanIQ could aid planners in generating treatment plans that push the limits of OAR sparing while maintaining routine clinical target coverage goals.

structures when the dose limit is just met, while others may choose to pursue further sparing after the dose limit is met. Automatic and knowledge-based planning aids have been developed with the aim of increasing plan quality while reducing variability.
The variability in plans is particularly true for complex treatment sites such as the head and neck. This region of the body has an abundance of important and radiation sensitive OARs and frequently requires treatments of irregularly shaped planning target volumes (PTVs) with potentially multiple dose prescriptions. One major area of concern for head and neck treatments is the dosimetric sparing of a patient's salivary glands, pharyngeal constrictors, and larynx. High doses of radiation to these organs can cause dry mouth (xerostomia) and difficulty swallowing (dysphagia). [2][3][4][5][6] Sparing a patient's salivary glands and larynx has been shown to reduce symptoms and increase patient quality of life. 2,[7][8][9] The QUANTEC review of dose-volume effects on salivary function by Deasy et al. concluded that for IMRT plans the mean dose to each parotid gland should be kept as low as possible. 3 It also states that, "a lower mean dose to the parotid gland usually results in better function, even for relatively low mean doses (<1000 cGy)." 3 The same review examining larynx and pharynx dosevolume effects had a similar conclusion stating that planners should minimize the volume of pharyngeal constrictors and larynx receiving 6000 cGy, and when, possible 5000 cGy. 6 Both publications emphasize the concept of minimizing the dose to these structures beyond the published/accepted benchmarks (i.e., as low as achievable). However, in practice, it is difficult to determine if a particular plan has in fact minimized the dose to these structures using a dose-volume histogram (DVH). The minimal dose to an OAR is predominantly dictated by the geometric relationship between the OAR and PTV(s). PlanIQ TM software (Sun Nuclear, Melbourne, Florida, USA) offers a tool called Feasibility DVH TM which quantitatively determines regions of a DVH that are impossible (red), difficult (orange), challenging (yellow), and probable (green) on a per OAR basis, based on an ideal dose falloff from the prescription dose at the target boundary. The impossible DVH (red) is defined as the DVH generated using the minimal dose that any voxel outside the targets must receive given 100% target coverage. We performed a study comparing salivary gland and larynx sparing with and without the use of PlanIQ's Feasibility DVH. We studied whether PlanIQ's Feasibility DVH could provide accurate estimates of OAR sparing and whether its use during treatment planning could facilitate increased sparing of patients' salivary glands and the larynx while maintaining target coverage and overall plan quality.

| MATERIALS AND METHODS
We identified 10 patients treated on one of two prospective protocols at our institution. All patients had primary lesions of the oropharynx and were node positive. Patients were originally treated using Tomotherapy (Accuray, Palo Alto, USA) (field width = 2.5 cm and pitch = 0.287-0.310) and retrospectively re-planned using Raystation (RaySearch Medical Laboratories AB, Stockholm, Sweden) on an Elekta Versa HD (Stockholm, Sweden) using volumetric arc therapy (VMAT) with three full 6 MV arcs (re-plan) and four full 6 MV arcs (IQ plan). The re-plan treatments were performed blinded to the results of the clinical plan and PlanIQ Feasibility DVH. For the IQ plan, the results from the PlanIQ Feasibility DVH analysis were available and used during treatment planning.
During planning, the MLC leaf motion was limited to 0.48 cm/degree. The PTVs and OARs used during clinical planning were the same as used during the re-plan and IQ plan. All OARs were expanded by 3 mm and PTVs pulled in skin 3 mm. A combination of equivalent uniform dose (EUD)-and DVH-based planning methods were used. The high risk PTV (PTV HR) was prescribed to 6000 cGy and the standard risk PTV (PTV SR) was prescribed 5400 cGy in 200 cGy fractions.
These two plans were delivered using simultaneous integrated boost (SIB). For the re-plan, a multi-criteria optimization (MCO) was generated for each patient followed by conversion to a deliverable plan and additional manual optimization. EUD optimization was used for sparing of OARs. Optimization structures were used for each OAR by subtracting the PTV SR with a 3 mm margin from the OAR. A maximum EUD (equation below) equal to 100-200 cGy less than the current average dose was used as a constraint with the "a" parameter equal to one prior to each round of optimizations.
where v i is the partial volume with absorbed dose D i .
This process was repeated until additional sparing was not achieved or sparing resulted in other plan issues such as hot spots, failing of clinical goals, etc. that could not be recovered. The goal of the re-plan was to maximally spare the contralateral parotid, contralateral submandibular gland, and larynx while still meeting our routinely used clinical goal sheet (Table 1). For the IQ plan, the Sun Nuclear PlanIQ Feasibility DVH information was made available during the planning process. The same (MCO) was performed; however, rather than iteratively reducing the optimization constraints for the OARs, the mean value derived from the impossible DVH was used as the criteria for the max EUD with the "a" parameter equal to one.
The EUD with the "a" parameter equal to one is equivalent to the mean dose. The re-plan was performed blinded to the results of the clinically delivered plan as well as the Feasibility DVH information from PlanIQ. The IQ plans were performed aware of the Feasibility DVH information. The IQ plans were not compared to the clinically delivered plan nor the re-plan during the planning process. A summary of the generated plans is shown in Table 2.
The Feasibility DVHs were generated by exporting the simulation

| RESULTS
All plans passed our institutional IMRT QA standard (range: 92.8%-98.6%). The comparison between the mean dose from the clinical, re-plan, and IQ plans to the impossible boundary from the Feasibility DVH (Impossible DVH) are shown in Fig. 1. The re-plans were able to provide increased sparing of OARs compared to the delivered plans and subsequently agreed better with the mean dose from Impossible DVH. The contralateral parotid and larynx were spared for all patients in both the clinically delivered plan and the re-plan.
On average, the re-plan reduced the dose compared to the clinical plan by approximately 750 cGy and 600 cGy for the contralateral parotid and larynx, respectively. For patients whose contralateral submandibular gland was spared in the clinical plan (7/10 patients), the re-plans reduced the mean dose by approximately 300 cGy.
The IQ plans were found to reduce the mean dose to the con- institution has subsequently has shifted to predominantly using Despite these credentials, we were surprised to see how much higher our OAR doses were compared to those predicted by PlanIQ.
Subsequently, the IQ plans were able to demonstrate that these predictions were quite accurate in terms of what could be delivered.
These results have led us to begin incorporating PlanIQ into our routine clinical planning processes.
In the future, our institution hopes to continue to utilize PlanIQ and to standardize methods by which this information is incorporated into the planning process. While we utilized 3 and 4-arc VMAT for this research, we have yet to determine if similar plans could be generated with few arcs in order to optimize clinical throughput. We also intend to investigate whether PlanIQ can not only improve plan quality but also if it may result in a reduction of planning time by providing reasonable estimates of expected DVHs upfront in the planning process. Evaluation of PlanIQ needs to be performed for sites other than head and neck.
Tools capable of providing predictions of what is dosimetrically achievable (and ideally optimal) are greatly needed in radiation treatment planning in order to reduce plan variability and ensure quality.
This work demonstrates for the first time that PlanIQ's Feasibility DVH agrees well with head and neck treatment plans that attempted to maximally spare salivary glands and the larynx. The addition of the Feasibility DVH information during planning led to an increased sparing of OARs compared to both clinical plans and plans blinded to this information. This suggests the Feasibility DVH could be a useful tool during planning and as a plan quality assurance tool. In the future, quantitative predictions such as the Feasibility DVH may be used in tandem with knowledge-based or auto-planning to provide the best of both worlds in terms of a tool that can address Feasibility, optimality, and deliverability. Additional studies are needed examining the incorporation of Feasibility DVHs during treatment planning and whether it could also lead to increases in clinical efficiency.

CONFLI CT OF INTEREST
Departmental research agreement with Sun Nuclear.