LEVEL SET PRIORS BASED APPROACH TO THE SEGMENTATION OF PROSTATE ULTRASOUND IMAGE USING GENETIC ALGORITHM

摘要

We propose a level set priors based approach to segment the prostate ultrasound image using the genetic algorithm (GA) optimization. Firstly, the ground truths are manually outlined in the training sets to generating the corresponding training level sets, and we derive the implicit boundary curve representation by using Principal Component Analysis (PCA). Secondly, a novel narrow band boundary feature is presented to determine the prostate edge. Thirdly, we use the genetic algorithm to optimize the parameters of the implicit curve representation. The experimental results demonstrate that the level set priors based method using genetic algorithm is robust and efficient.