Algoritmo transgenético para os problemas da geometria e da intensidade em IMRT

• Ano de defesa: 2021
• Autor(a) principal: Cunha Neto, Luís Tertulino da
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Rio Grande do Norte; Brasil; UFRN; PROGRAMA DE PÓS-GRADUAÇÃO EM SISTEMAS E COMPUTAÇÃO
• Palavras-chave em Português: Radioterapia; IMRT; Problema da Geometria; Problema da intensidade; Algoritmo transgenético; Épsilon-restrito; Funções de dose-volume
• Link de acesso: https://repositorio.ufrn.br/handle/123456789/32619

Intensity Modulated Radiotherapy (IMRT) is a form of treatment of cancerous diseases in which the patient is irradiated with radiation beams, aiming to eliminate tumor cells while sparing healthy organs and tissues as much as possible. Each beam is divided into beamlets that may emit different radiation doses. A treatment plan is composed of: (a) a set of beam directions (angles); (b) the amount of radiation emitted by the beamlets of each beam; and (c), a radiation delivery sequence. The elaboration of a plan can be modeled by optimization problems, usually NP-hard, where steps (a), (b) and (c) are called Geometry, Intensity (or Fluence Map) and Realization problems, respectively. This work addresses the first two. An evolutionary algorithm is proposed for the joint solution of these two problems; namely: an hybrid Transgenetic Algorithm. It uses an adaptation of the -constraint method to compute the fluence map of a set of beams. Linear and quadratic approximation functions are proposed for a particular type of (non-convex) function present in radiotherapy optimization: the dose-volume function. Two groups of computational experiments are carried out, using real cases of liver cancer, to ascertain the algorithm effectiveness: the first one with the tumor’s dose as constraint, and the second one with the tumor’s dose as objective function. The results of the objective functions show that the second technique is more appropriate to achieve better doses in the tumor. Other results regarding the effectiveness of the approximation functions and the components of the Transgenetic Algorithm are also discussed.