APPLYING OPERATIONS RESEARCH MODELS TO PROBLEMS IN HEALTH CARE

Abstract

Intensity- modulated radiation therapy is a form of cancer treatment that directs high energy x-rays to irradiate a tumor volume. In order to minimize the damage to surround-ing tissue the radiation is delivered from multiple angles. The selection of angles is an NP-hard problem and is currently done manually in most hospitals. We use previously evaluated treatment plans to train a machine learning model to sort potential treatment plans. By sorting potential treatment plans we can find better solutions while only evalu-ating a fifth as many plans. We then construct a genetic algorithm and use our machine learning models to search the space of all potential treatment plans to suggest a potential best plan. Using the genetic algorithm we are able to find plans 2% better on average than the previously best known plans.

Proton therapy is a new form of radiation therapy. We simulated a proton therapy treatment center in order to optimize patient throughput and minimize patient wait time. We are able to schedule patients reducing wait times between 20% and 35% depending on patient tardiness and absenteeism.

Finally, we analyzed the impact of operations research on the treatment of pros-tate cancer. We reviewed the work that has been published in both operations research and medical journals, seeing how it has impacted policy and doctor recommendations.