3D Prostate Boundary Reconstruction from 2D TRUS Images

The shape, volume and location of the prostate are essentially important information in prostate cancer diagnosis and treatment, e.g. ultrasound-guided biopsy and brachytherapy. In this paper, 3D boundary of the prostate is reconstructed from a series of 2D transrectal ultrasound (TRUS) images. Manual outlining of the prostate is tedious, time-consuming, and greatly dependent on the operator's experience, so a semi-automatic segmentation method is used to generate 2D contours of the prostate using a modified discrete dynamic contour (DDC) method. Contour-based surface reconstruction technique is employed to construct 3D surface of the prostate. The performance of modified DDC method is evaluated by comparing the semi-automatically segmented contours with those outlined by an expert with distance-based and area-based metrics. The average distance between the semi-automatically and manually outlined contours is less than 3 pixels ~ 0.44 mm. 3D surface of the prostate from 2D TRUS images is visualized using OpenGL technique.