Uma abordagem multiobjetivo para o problema da realização da radioterapia de intensidade modulada

Abstract

Algorithms are an essential part of the radiation therapy planning, and under a optimization point of view, can be divided in three subproblems. Defining the angles by which radiation will be shot and prescribe a fluence map for each angle are two of them. This work investigates the third problem, called realization problem. It consists on defining a sequence of configurations for a device (called multileaf collimator) which correctly delivers the prescribe doses to the patient. A common model for this problem is the decomposition of a matrix in a weighted sum of (0-1)-matrices, called segments, whose rows only have consecutive ones. Each segment represents a setup of the collimator. Other constraints can be also considered. The realization problem has three objectives. The first one is to minimize the sum of the weights associated to the segments. The second is to minimize the number of segments. The third minimizes the movement of the leaves. This work investigates and present algorithms for two variants of the problem: unconstrained and constrained. A new greedy and randomized algorithm – GRA – was developed firstly for the unconstrained variant and then extended for the constrained variant. Its results was compared to other algorithms from the literature, under mono and multiobjective points of view. On the unconstrained problem, experiments show that GRA outperforms the other algorithms by all measured indicators. On the constrained problem, GRA presented competitive results, specially on the second objective, in which it presented the best results.