A column generation-based algorithm for Volumetric Modulated Arc Therapy (VMAT) treatment plan optimization
|University of Michigan
|Mathematics & Computer Building (MC) 5158
External beam radiation therapy is a common treatment for many types of cancer. During such treatment, radiation is delivered with a gantry, equipped with a radiation source, that is pointed at the patient from various angles. Optimization models are commonly used in individualized treatment planning, and formulation and solution methods for such models have been an area of active research and collaborations. VMAT is a particular technique for delivering radiation, in which the gantry continuously rotates around the patient while the leaves of a multi-leaf collimator (MLC) move in and out of the radiation field to "shape" it. Proposed over a decade ago, this technique has the potential to produce treatments of high quality similar to, e.g., Intensity Modulated Radiation Therapy (IMRT), but requiring less time for treatment delivery. Recently, treatment systems capable of delivering VMAT treatments became commercially available, necessitating the development of relevant treatment planning methods. We propose one such method, which uses optimization models and column generation-based heuristics to produce high-quality VMAT treatment plans that allow for (i) dynamically adjustable gantry speed; (ii) dynamically adjustable dose rate; and (iii) MLC leaf speed constraints.
This talk is based on joint work with Fei Peng and H. Edwin Romeijn at the University of Michigan and Xun Jia, Xuejun Gu and Steve B. Jiang at the University of California San Diego.
Marina Epelman received her PhD in Operations Research from Massachusetts Institute of Technology (MIT) in 1999. The same year she joined the Industrial and Operations Engineering department at the University of Michigan, where she is now an associate professor. Her research interests lie in developing optimization models and algorithms for problems arising in oncological treatment planning, transportation, staffing, manufacturing and other applications, solved by techniques of convex optimization, sampling-based algorithms, dynamic programming and stochastic optimization.