Comparison of the progressive resolution optimizer and photon optimizer in VMAT optimization for stereotactic treatments

Abstract The photon optimization (PO) algorithm was recently released by Varian Medical Systems to improve volumetric modulated arc therapy (VMAT) optimization within Eclipse (Version 13.5). The purpose of this study is to compare the PO algorithm with its predecessor, progressive resolution optimizer (PRO) for lung SBRT and brain SRS treatments. A total of 30 patients were selected retrospectively. Previously, all the plans were generated with the PRO algorithm within Eclipse Version 13.6. In the new version of PO algorithm (Version 15), dynamic conformal arcs (DCA) were first conformed to the target, then VMAT inverse planning was performed to achieve the desired dose distributions. PTV coverages were forced to be identical for the same patient for a fair comparison. SBRT plan quality was assessed based on selected dose–volume parameters, including the conformity index, V 20 for lung, V 30 Gy for chest wall, and D 0.035 cc for other critical organs. SRS plan quality was evaluated based on the conformity index and normal tissue volumes encompassed by the 12 and 6 Gy isodose lines (V 12 and V 6). The modulation complexity score (MCS) was used to compare plan complexity of two algorithms. No statistically significant differences between the PRO and PO algorithms were found for any of the dosimetric parameters studied, which indicates both algorithms produce comparable plan quality. Significant improvements in the gamma passing rate (increased from 97.0% to 99.2% for SBRT and 96.1% to 98.4% for SRS), MCS (average increase of 0.15 for SBRT and 0.10 for SRS), and delivery efficiency (MU reduction of 29.8% for SBRT and 28.3% for SRS) were found for the PO algorithm. MCS showed a strong correlation with the gamma passing rate, and an inverse correlation with total MUs used. The PO algorithm offers comparable plan quality to the PRO, while minimizing MLC complexity, thereby improving the delivery efficiency and accuracy.


| INTRODUCTION
In the past decade, stereotactic radiation therapy has been increasingly used in the management of both intracranial (stereotactic radiosurgery, SRS) and extracranial tumors (stereotactic body radiation therapy, SBRT). This technique is able to precisely deliver very high doses to the tumors while sparing the adjacent normal tissues in just a few fractions, and has been proven to provide superior or comparable treatment outcomes and is cost-effective relative to alternative conventional techniques. 1 Volumetric modulated arc therapy was first introduced and implemented into the clinic as a novel radiation delivery technique 11,12 that was a variation on static field IMRT. Unlike DCA, VMAT plans are generated by an inverse planning algorithm, which allows modulation of the gantry rotation speed, dose rate, and also the position and speed of MLC. It is recommended that VMAT plans always be validated with a patient-specific QA measurement before treatment. Compared to VMAT, DCA techniques offer advantages in terms of reduced multileaf collimator (MLC) motion complexity, positioning error, and delivery efficiency, thus DCA may be the desired delivery modality from a technical treatment delivery viewpoint. However, VMAT may be favored due to its increased ability for dose shaping when certain critical organs sparing becomes important.
The photon optimization (PO) algorithm was recently released by Varian Medical System (Palo Alto, CA, USA) to improve IMRT and VMAT optimizations in Eclipse treatment planning system (TPS) version 13.5. The PO algorithm combines the dose-volume optimizer (DVO) used for static field IMRT and the progressive resolution optimizer (PRO) used for VMAT plans from the previous Eclipse version.
Both PRO and PO algorithms create VMAT plans based on dosevolume objectives, and generate a sequence of control points, which define MLC leaf positions and MU/degree as a function of gantry angle. For both PRO and PO algorithms, the initial conditions are defined using control points to represent each VMAT field; a multiresolution approach and an objective function (the sum of the dose-volume and user-defined objectives) are used to optimize the plan. The multiresolution dose calculation (MRDC) algorithm is used to increase the dose calculation accuracy with progressive dose calculation segments using a point cloud-based model. 13 The optimization process goes through four multiresolution level, in which number of dose calculation segments increase progressively at each level. The number of control points keeps unchanged during the whole optimization process.    Each clinical plan was normalized such that at least 95% of the PTV received the prescription dose, and more than 99% of the PTV received at least 90% of the prescription dose. The maximum point dose and dose-volume constraints of several critical organs are listed in Table 1 for both 54 and 50 Gy protocols. [14][15][16] All SRS patients were immobilized in BrainLab SRS masks (Brain-Lab AG, Feldkirchen, Germany) and received IV contrast prior to the CT scan when possible. All patients also underwent MR scans, which were registered with the planning CT to define the gross tumor volume (GTV). The GTV was expanded by 1 mm margin uniformly to form the PTV. MR scans were acquired on a GE Signa EXCITE 3.0T MRI scanner (GE Healthcare, Waukesha, WI, USA) with 1 mm slice thickness and a field of view large enough to include the entire surface of the patient. Postcontrast T1-weighted axial scans were used for all cases. 6X-FFF beams with 3-4 table kicks were used for the original clinical VMAT plans, and at least 99% of PTV and 100% GTV coverages were enforced. The maximum point dose and dosevolume constraints for critical structures followed the guidelines of AAPM TG-101. 14 All the clinical plans were optimized in Eclipse version 13.6 with the PRO algorithm, which is based on dose-volume objectives. All plans were reoptimized with the new PO algorithm in Eclipse version 15, which offers a VMAT optimization based on an initial DCA plan as described above. The PO algorithm determines the optimal field shape and intensity by interactively conforming the dose distribution to the desired objectives until an optimum solution is reached. In the new PO algorithm, DCA was first used to conform to the target and followed by VMAT inverse planning to achieve the desired dose distribution.
The DCA beams were automatically fit with 3-4 mm margin around the PTV, and the same beam configurations (gantry, collimator and table angles) were used in the re-plan. In order to isolate the variation due to the planning optimization algorithm, the same planning objectives, constraints, and weightings for the target and critical organs were used for both the PRO and PO plans. The settings were also retained for normal tissue optimization (NTO) and the MU objectives.
PTV coverage was forced to be identical for the same patient in order to have a comparison between the PRO and PO algorithms (at least 95% for lung SBRT   Better QA results from PO plans can be attributed to reduced plan modulation (MCS and AA) for the new PO algorithm, which is consistent with previous a finding. 18 Figure 4  where N is the number of moving leaves and pos n is the position of leaf n.
The AAV is based on the area defined by opposing MLC leaves in a CP normalized to the maximum area in the arc, which is defined by the maximum apertures for all leaf pairs over all CPs in the arc AAV CP ¼ P A a¼1 hpos a i left bank À hpos a i right bank P A a¼1 hmaxðpos a Þi left bank2arc À hmaxðpos a Þi right bank2arc ; where A is the number of moving leaves in the arc.
The MLCs are continuously moving between CP while MU are being delivered, so MCS considers the mean values of LSV CP and AAV CP between adjacent CPs. This product is weighted by the percentage of total MU delivered between the adjacent CPs: where I is the number of CP and MU CPi,i+1 are the MU delivered between two consecutive CPs.
The AA term is defined using the notation from above as hpos a i left bank À hpos a i right bank Â w a where w a is the width of leaf a AA arc ¼ MUarc h i I À 1 :