Segmentation and reconstruction of prostate from transrectal ultrasound images

In this paper, the 3D shading image of prostate is reconstructed from a series of transrectal ultrasound images. Each slice image is first smoothed by direction-based low pass filtering. The center of prostate is estimated and then a search process radiated from the estimated center is performed to locate the prostate boundary based on matched filtering. After this step, edge linking including a priori knowledge is employed to obtain a 2D contour. The Hermite-surface approach is used to interpolate the spatial data to reconstruct 3D object's surface from a series of 2D contours. The diffuse reflection and Gouraud shading are used to create the depth effect of the reconstructed 3D image. The 3D spatial structure information obtained about prostate is very contributive to physicians in the clinical diagnosis or surgical applications.