IMRT Inverse Planning Using Simulated Annealing

Abstract

Radiation therapy is an important type of cancer treatment, being delivered to more than half of all cancer patients. Intensity Modulated Radiation Therapy (IMRT) is one type of radiation therapy where it is possible not only to conform the dose distribution to the shape of the volumes to treat, but also to have different radiation intensity profiles that will allow a better coverage of the volumes to treat and a better sparing of all other structures. Planning an IMRT treatment is a time-demanding task, based on a trial and error procedure. We propose an inverse planning approach to this problem based on simulated annealing (SA) and a dynamic neighborhood concept. The planning of an IMRT treatment can be interpreted as having several stages, from choosing proper incidence directions until deciding on the movement of the multiple collimator leaves that will produce the desired intensity profiles. We consider here mainly the first of these stages: the problem of choosing the best intensity directions. This problem is known to be highly non-convex and with many local minima. SA is a well known metaheuristic that has proven to be capable of escaping local minima. Some preliminary computational results using ten clinical examples of already treated patient cases of head-and-neck tumors at the Portuguese Institute of Oncology of Coimbra (IPOC) will be described.