According to the World Health Organization reports, cardiovascular diseases are the worldwide leading cause of death. However, they can be controlled using proper diagnosis tools and treatments. In this way, the diagnosis assisted technological tools are important to reduce the number of deaths caused by this type of diseases. This dissertation assess optimization methods (Differential Evolution, Generalized Pattern Search, Mesh Adaptive Direct Search, Nelder-Mead Algortihm) applied in the search of parameters that model blood vessels structures in order to improve the results and processing time of a vascular tree segmentation method in medical images, as proposed in (Oliveira, 2013). In this work, anatomical concepts and the characteristics of the images used in this study are presented. The optimization methods assessed and the methodology for the segmentation of the vascular tree, whose parameters are to be optimized, are described. Based on that, the procedure to assess different optimization methods is formulated through a quantitative analysis using a hypothesis test formulation of difference between paired means. This test evaluates the performance of the optimization methods using randomly chosen samples in a computerized tomography exams database composed by synthetic, coronary, hepatic and nervous fiber of the olfactory system images. According to the hypothesis test results, the optimization method with the best performance, both in accuracy and computational cost, is selected and the conclusions of this work are elaborated as well as the proposals for further research.