Enhanced graph cuts for brain tumor segmentation using bayesian optimization

Abstract

Brain tumor segmentation is a difficult task, due to the shape variability that malignancy brain structures exhibit between patients. The main problem in this process is that the tumor contour is usually computed from parametric models that need to be well-tuned to perform an accurate segmentation. In this paper, we propose an enhanced Graph cut on which the model parameters are selected through a probabilistic approach. Here, we use Bayesian optimization to find the optimal hyperparameters that segment the tumor volume accurately. The experimental results show that by using Bayesian optimization, the graph cut model performs an accurate segmentation over brain volumes in comparison with common segmentation methods in the state-of-the-art.