Direct Aperture Optimization for Intensity Modulated Radiation Therapy Treatment Planning
Traditionally, IMRT treatment planning has been performed in two phases: a first phase, called
fluence map optimization, in which optimal beamlet intensities are determined, followed by a second, leaf sequencing, phase in which these fluence maps are decomposed into deliverable apertures and corresponding intensities. Most approaches that integrate these two phases by simultaneously identifying high quality apertures and intensities are heuristic in nature, mainly due to the astronomical number of available apertures. We instead formulate an integrated model that is exact in the case where only convex treatment plan evaluation criteria are considered. Examples of such criteria are quadratic voxel-based penalties, EUD, and tail averages (CvaR) in the objective function, and constraints on measures such as TCP and NTCP. This approach is based on successively adding the most promising apertures to a relatively small so-called segment-weight optimization formulation of the problem. We find empirically that this approach yields high-quality treatment plans with only a relatively small number of apertures. Moreover, this approach lends itself very well to incorporating delivery aspects that cannot be incorporated in a beamlet-based fluence map optimization approach – such as transmission and the tongue-and-groove effect. In addition, beam-on-time can be incorporated as an explicit criterion or constraint. We can accommodate various aperture delivery constraints, such as row-convexity, interdigitation, connectedness, and jaws-only delivery. We will end the talk with current and future research directions.
Edwin Romeijn received his Ph.D. in 1992 from Erasmus University Rotterdam in The Netherlands. Since then, he has been a faculty member at the Rotterdam School of Management, the University of Florida and, since 2008, the University of Michigan. In addition, he has been a visiting faculty member at Columbia University, Maastricht University, National University of Singapore, University of California at Berkeley, University of Pittsburgh, and Massachussetts Institute of Technology. He has taught courses in operations research, stochastic processes, optimization, applied probability and statistics, financial engineering, supply chain management, and decision support systems. His research focuses on optimization theory and applications, in particular in the areas of supply chain optimization and optimization in health care. He is the author of more than seventy-five peer reviewed journal publications.