As an honors undergraduate in industrial engineering, Troy Long focused on developing algorithms for scheduling in semiconductor facilities. Now three years into doctoral studies in industrial and operations engineering at the University of Michigan, Troy has turned to a very different application that may help doctors more efficiently determine effective radiation treatment plans for cancer patients. “I wanted to do something with health care and large-scale optimization,” Troy said. “I thought I might be able to make a difference.”
Troy returned to campus recently to present his research and share advice on graduate school and research funding with students and professors in the industrial engineering department. His talk, “Beam Orientation Optimization for Radiation Therapy Treatment Planning,” presented research led by his UM mentor Edwin Romeijn.
“The goal is to develop a high quality treatment plan: one that delivers a prescribed dose to the cancer targets, but spares critical structures and healthy tissue,” Troy explained. Treatment delivery uses a linear accelerator fitted with a multileaf collimator to shape the beam’s output. Visualize a patient stretched out on a platform, while a massive machine moves around him or her, shooting radiation beams into the patient to dose the affected area. Currently, radiation oncologists select the beams’ locations heavily based on intuition and experience. “You know you want to give the optimal distribution of dose – radiation – to the patient. What you don’t know is how to position the machine,” Troy said.
Troy’s team is collaborating with medical physicists and radiation oncologists to develop an integrated model that explicitly considers location of the beam, the shapes of the apertures created by the MLC’s sliding tungsten leaves, and the intensity of the radiation passing through the apertures. “Basically, we’re looking at ways to quantify the quality of the beam that could be added to the model that is used to treat the affected area,” he said.
Troy is doing an extensive amount of coding to support this research and is applying these techniques to clinical cases. He is starting work in parallel algorithms, which can be executed in several pieces at a time on multiple processors, to make the planning process more efficient. He’s published one paper and co-authored another, and this year received an NSF Graduate Research Fellowship. “I want to develop a good optimization toolbox – because at the core, engineering is about problem solving,” Troy said. He’s considering a host of options, post-Ph.D., from industry to academia. “I see myself as a jack-of-all-trades, and I plan to apply for jobs everywhere,” he said.