Using Dynamic Discrete Contours for segmentation of prostate ultrasound images in two dimensions

Prostate cancer is the second cancer type, after skin cancer, which most commonly affects men around the world. It also has the second highest mortality rate after lung cancer in Perú. Segmentation of the prostate boundary from ultrasound images is one of the most important tasks for diagnosis and treatment. Different algorithms of automatic segmentation have been created without much success. As a result, nowadays this task is performed manually, becoming an arduous, time-consuming and heavily user dependent job due to limited quality of ultrasound images. This works presents a short review of methods that have been proposed for semiautomatic segmentation, and implements a segmentation algorithm based on Discrete Dynamic Contours, which have been previously shown to have good results in this task. For this implementation, initialization requires selection of 4 points which will not change their positions in order to delimit the prostate location. Pre-processing is used to improve contrast quality and reduce noise, using Sticks and Anisotropic Diffusion algorithms. Results show accuracy and sensitivity over 90% in the segmentation of prostate in two ultrasound volumes.