Global convergence analysis of fast multiobjective gradient-based dose optimization algorithms for high-dose-rate brachytherapy.

摘要: We consider the problem of global convergence gradient-based optimization algorithms for interstitial high-dose-rate (HDR) brachytherapy dose using variance-based objectives. Possible local minima could lead to only sub-optimal solutions. perform a configuration space analysis representative set entire non-dominated solution space. A three prostate implants is used in this study. compare results obtained by conjugate gradient algorithms, two variable metric and fast-simulated annealing. For algorithm BFGS from numerical recipes, large fluctuations are observed. The limited memory L-BFGS FRPR globally convergent. Local or degenerate states not study possibility obtaining solutions optimal rearrangement warm start mechanism. surface volume variance their derivatives, method proposed which significantly reduces number required operations. time, ignoring preprocessing step, independent sampling points planning target volume. Multiobjective HDR new modified computation objectives derivatives has been accelerated, depending on points, factor range 10-100.