Parameter optimization in HN-IMRT for Elekta linacs.

Abstract

Planning and delivery in HN-IMRT has been challenging for the Elekta linac because of numerous machine limitations. Direct aperture optimization (DAO) algorithms have had success in simplifying the planning process and improving plan quality. Commercial adaptations of DAO allow for widespread use in many clinics; however clinical validation of these methods is still needed. In this work we evaluated Pinnacle3 commercial software for HN-IMRT on the Elekta linac. The purpose was to find a set of planning parameters that are applicable to most patients and optimal in terms of plan quality, delivery efficiency, and dosimetric accuracy. Four types of plans were created for each of 12 patients: ideal fluence optimization (FO), conventional two-step optimization (TS), segment weight optimization (SW), and direct machine parameter optimization (DMPO). Maximum number of segments (NS) and minimum segment area (MSA) were varied in DMPO. Results showed DMPO plans have the best optimization scores and dosimetric indices, and the most consistent IMRT output among patients. At larger NS (> or = 80), plan quality decreases with increasing MSA as expected, except for MSA<8 cm(2), suggesting presence of local minima in DMPO. Segment area and MUs can vary significantly between optimization methods and parameter settings; however, the quantity 'integral MU' remains constant. Irradiation time is linearly proportional to total plan segments, weakly dependent on MUs and independent of MSA. Dosimetric accuracy is independent of DMPO parameters. The superior quality of DMPO makes it the choice for HN-IMRT on Elekta linacs and its consistency allows development of 'class solutions'. However, planners should be aware of the local minima issue when pushing parameters to the limit such as NS<80 and MSA<8 cm(2). The optimal set of parameters should be chosen to balance plan quality and delivery efficiency based on a systematic evaluation of the planning technique and system constraints.