Toward the combined optimization of dynamic axes (CODA) for stereotactic radiotherapy and radiosurgery using fixed couch trajectories

Abstract

Purpose To develop a novel system for patient‐specific combined optimization of couch, collimator, and gantry angles for use in volumetric modulated arc therapy (VMAT) treatment planning. The system was designed to produce highly compact dose distributions by extensively sampling the 4π space. Automated fixed couch trajectory planning was used to reduce normal tissue doses by avoiding beams‐eye‐view (BEV) overlap with organs‐at‐risk (OARs) and improve monitor unit (MU) efficiency through collimator angle optimization. Methods By merging distinct BEV objective functions used to optimize the couch rotation angle and collimator angle, a three‐dimensional (3D) cost space (the CODA cube) was constructed with axes of gantry, couch, and collimator rotation angles. At each voxel in this CODA cube, the cost of implementing this combination of axes positions in fixed couch trajectories was quantified. The CODA cube was sampled and explored using a modified constrained Bellman‐Ford algorithm to suggest low‐cost fixed candidate arcs on each plane of the space, from which 10‐arcs are chosen throughout the 3D space using a k‐means clustering algorithm. These fixed couch trajectories were then imported into the Eclipse treatment planning system (v.11) and inverse‐optimized according to clinical standards. Eight artificial cranial targets were contoured in a test‐patient anatomy, and seven treatment plans were generated from combinations of three and four targets. The CODA cube optimized plans were compared to standard 4‐arc VMAT plans for cranial stereotactic radiotherapy/surgery that were optimized for the same sets of targets; maximum dose to each OAR, V12Gy to normal brain, conformity, and total MUs were compared. Both planning methods were inverse‐optimized with identical dosimetric objectives. Results CODA plans resulted in a reduction in maximum dose to OARs of 20.6% (P < 0.01), with maximum brainstem dose decreased by 2.63 Gy (P = 0.031) on average when compared to the standard arc arrangement. The mean reduction in total MU was 8.6% (P = 0.156), the mean increase in the inverse of the van’t Riet conformation number was 0.1%, (P = 0.67) and the mean decrease in normal brain tissue receiving 12 Gy or higher was 3.9% (P = 0.16), when compared to the standard VMAT arc configuration (n = 7). Conclusions The optimization of couch, collimator, and gantry angles simultaneously using a 3D optimization space achieved improvement on multiple clinical metrics when compared to the standard VMAT arc configuration. A statistically significant sparing to OAR maximum doses was seen. Combining these optimizations may yield superior results to independent optimization.

SBRT of ventricular tachycardia using 4pi optimized trajectories

“…The method of a dynamic trajectory‐based couch motion proposed by MacDonald et al. 33 , 34 was used in this work for generating treatment arc geometries with optimized trajectories that minimize the geometric overlap between PTV and OARs in the BEV. The 4pi algorithm was implemented using the Eclipse Scripting API (ESAPI), thus allowing for definition of optimized trajectories within the Eclipse treatment planning system.…”

“…Figure 2(c) shows an overlap map calculated for one patient and for selected OARs with the final optimized trajectories superimposed as straight lines, and dark maroon patches representing the forbidden collision zones. The ESAPI application provides the final optimized set of allowable minimumcost arcs as described by MacDonald et al 34 A treatment plan is then created within Eclipse using the set of optimized trajectories (couch and gantry angles) generated by the ESAPI. Figure 2(d) shows the corresponding arcs planned in Eclipse with the optimized trajectories.…”

“…[29][30][31][32][33][34] For example, MacDonald et al proposed a customizable algorithm in order to generate a navigable ideal trajectory based on the amount of overlap in the beamseye-view (BEV) between the planning target volumes PTVs and OARs, and thus demonstrating the capacity to improve radiation therapy dose distributions in cranial SRT. 33,34 Their results show an average decrease of 19% and 15% in mean and maximum doses respectively to OARs when compared to conventional treatment trajectories. 33 It is well recognized that 4pi methodology has already been extensively explored, but mainly for stereotactic techniques for treating brain lesions, to minimize exposure of healthy tissues.…”

“…Methods of optimizing this geometry in an automated way have been proposed in the literature. 29 , 30 , 31 , 32 , 33 , 34 For example, MacDonald et al proposed a customizable algorithm in order to generate a navigable ideal trajectory based on the amount of overlap in the beams‐eye‐view (BEV) between the planning target volumes PTVs and OARs, and thus demonstrating the capacity to improve radiation therapy dose distributions in cranial SRT. 33 , 34 Their results show an average decrease of 19% and 15% in mean and maximum doses respectively to OARs when compared to conventional treatment trajectories.…”

Technical Note: Collimator angle optimization for multiple brain metastases in dynamic conformal arc treatment planning

“…They address delivery time for sweep-style delivery rather than exposed non-target area. A study that addresses the exposed non-target area is the automated fixed couch trajectory method for VMAT proposed by MacDonald et al 11 They construct a three-dimensional cost space cube based on the gantry, couch, and collimator angles, where each voxel is associated to a cost which is the combination of the magnitude of organs at risk present in the beam's eye view and the exposed area of non-target anatomy. Arcs where the total gantry travel length is greater or equal than 30 degrees and arcs which are feasible with respect to machine constraints K E Y W O R D S collimator angle optimization, dynamic collimator trajectory, dynamic conformal arcs, multiple metastases, treatment planning F I G U R E 1 Comparison of the beam's eye view for the same control point at non-optimal collimator angle (a) and at optimal collimator angle (b).…”