Introduction and Objectives
3D prostate gland depiction from Transrectal Ultrasound images assists the clinician in several situations such as accurate gland volumetry, tumor margin estimation, PSA density calculation, 3D visualization, Brachytherapy, HIFU, Cryosurgery etc.
Material and Methods
Our method involves preprocessing (edge preservation, noise reduction, smoothing) and prostate gland segmentation. The proposed algorithm uses the deformable model (snake), a method that involves designing an energy function and then optimizing this function. The initial contour or the seed points were estimated automatically from the image with the highest diameter both in transversal and sagital plane. Special measures were taken to deal the high speckle noises and complex shapes of prostate boundaries. In general a series of imaging enhancement methods were used such as histogram thresholding, median filter, stick's filter and regional contrast enhancement. The 2D contour was extracted from approximately 20 2D images, which were processed with the help of an image editor in order to reconstruct the 3D of the prostate gland.
The proposed method obtained excellent gland segmentation result with average overlapping areas of 91%, as it was compared with expert radiologist's segmented images. It is possible to segment the gland from the interior and to identify tumourous lesions.
The proposed algorithm requires no user interaction, provides excellent segmentation results and can be applied to several diagnostic and therapeutic modes.
© 2009 European Association of Urology. Published by Elsevier Inc. All rights reserved.