Beam Selection For Mlc-Based Robotic Radiation Therapy
bo wang,juan wang,jie li,j fan,jiuhong kang,cuiqing ma
DOI: https://doi.org/10.1016/j.ijrobp.2013.06.2015
2013-01-01
Abstract:Recent studies have shown that the robotic radiosurgery system equipped with a multileaf collimator (MLC) is favorable for treatment time reduction and plan quality improvement, comparing with the existing cone-based systems. Advanced beam selection and treatment optimization methods are needed to fully realize the potential of this new radiation therapy system. The purpose of this study is to develop a novel geometry-based automated beam selection program to facilitate advanced treatment optimization for MLC-based robotic radiation therapy. Our candidate beam set consists of 94 non-coplanar beams, each defined by a vector connecting a robotic radiosurgery node from a chosen body path set and a target point. The MLC leaves are configured to approximate a circle with radius R at 80 cm SSD, such that the beam cross section fully covers the planning target volume (PTV), which also covers certain area on patient surface. It is known that to deliver a certain dose to PTV the larger the total covered surface area is, the lower the surface dose will be and that the more beams are used in a plan the longer the treatment delivery will be. As a compromise, we aim to choose an adequate number of beams that will cover as much surface area as possible. By setting the MLC opening and constraint on the degree of overlap, we have selected various numbers of beams for different treatment sites. To evaluate the selection results, we generated IMRT plans with the selected beams using the TPS and compared them with conventional IMRT and VMAT plans, which were generated based on our current clinical requirements and planning strategies. We used eight clinical cases including prostate, pancreas and spine treatments. Our results show that we can cover 88.5% of the total number of surface voxels (NSV) that are covered by all the 94 beams with 22 to 29 beams for R = 4 cm with a 20% maximum degree of overlap (DO). The beam number is reduced to 15 to 20 to cover 84.7% of NSV for R = 5 cm with a 20% DO, and 18 to 22 to cover 78.8% of NSV for R = 4 cm with a 10% DO. All the IMRT plans based on the selected beams provide conformal dose coverage for the PTV. The plan quality, in terms of critical structure sparing and PTV dose homogeneity is better than or at least equal to those of conventional IMRT and RapidArc plans for the cases studied. Taking a pancreas case as example, with 13 selected beams we obtained a plan for which the PTV Dmax, Dmin, D98%, D2% and Dmean (doses are normalized so that 95% PTV is covered by 100% prescription) are 109%, 87%, 98%, 106% and 103% respectively. Dmean to the liver, left and right kidneys are 19%, 11%, and 18% respectively, while Dmax to the spinal cord is 30%. With our method, one can efficiently select a small but effective set of non-coplanar beams. Our program can be used in combination with any optimization algorithm to generate high-quality plans for MLC-based robotic radiation therapy.