ABSTRACT
This chapter proposes a level-set-based framework for the delineation of the prostate from three-dimensional (3D) diffusion-weighted (DW) magnetic resonance imaging (MRI). The level-set deformable model is guided by a stochastic speed function that is derived using nonnegative matrix factorization (NMF), which extracts meaningful features from a large dimensional feature space. The NMF attributes are calculated using information from the DW-MRI intensity, a probabilistic shape model, and the spatial interactions between prostate voxels. The shape model is built using a set of training prostate volumes and is then updated during the process of the segmentation using an appearance-based approach that considers both a voxel’s location and its intensity value. The pairwise spatial interactions are modeled using a second-order 3D Markov-Gibbs random field (MGRF). Experiments show that using this information along with NMF-based feature fusion to guide the evolution of the level-set increases the segmentation accuracy compared with previously proposed methods using two metrics, the Dice similarity coefficient (DSC) and Hausdorff distance (HD). The proposed approach achieved an average DSC of 0.870 ± 0.03 and an average HD of 5.72 ± 2.35 mm3 compared to an average DSC of 0.833 ± 0.07 and an average HD of 6.74 ± 2.04 for a maximum a posteriori (MAP)-based level-set method, and an average DSC of 0.810 ± 0.05 and an average HD of 9.07 ± 1.64 for a level-set driven only by intensity and shape information.
