Brachytherapy cancer treatment optimization using simulated annealing and artificial neural networks

This paper presents research aimed at improving brachytherapy cancer treatments. The focus of the research is to optimize the locations of the applicators used in brachytherapy treatment plans using artificial intelligence. Currently the optimization of the applicators occurs before the treatment is carried out due to the lengthy optimization process. This work investigates the possibility of using artificial neural networks (ANNs) to overcome this difficult. The reasons for using an ANN are the speed and generalization abilities it can possess. Using a single hidden layer backpropagation ANN we have been able to optimize applicator positions in 2D square tumours up to 3 cm in cross sectional size in less than 1 second. These results are more than 300 times faster than the next fastest method. Using our ANN optimization method we would be able to optimize a treatment after each applicator is inserted.