Various attempts have been made to improve prostate visibility. Daanen et al. (1) attempted to fuse MRI data with TRUS data for more reliable

image processing and prostate volume identification. However, MRI is not part of the standard of care for prostate radiation treatment and would add expense and time to the treatment. Furthermore, because TRUS and MRI are carried out under different conditions (different rectum deformation check details by endorectal coils or TRUS and different leg and pelvis position), complex deformable registration techniques must be used. In previous reports by Sahba et al. (2) and Pathak et al. (3), image processing techniques have been used for ultrasound image enhancement. Different imaging modalities lead to different segmentation results. For example, Smith et al. (4) have evaluated the reproducibility BMN 673 concentration and modality differences of prostate contouring, after brachytherapy implants, using three-dimensional (3D) TRUS and T2-weighted MR and CT imaging. Prostates from 10 patients with early-stage prostate cancer (T2b or less) were segmented twice by seven observers. Their results showed high contouring

variability of the anterior base and apex in 3DTRUS, whereas the prostate–rectum interface had the smallest variability. In TRUS imaging, the interobserver variability of prostate contouring is high. A study by Choi et al. (5) showed that prostate volume measurement by TRUS may vary among observers when patients have large prostates (≥30 cm3). The average volume difference between 101 prostates measured by two experienced observers was reported as 6.00 cm3 for prostates with a

mean measured volume of 30 cm3 or more and Etofibrate 1.51 cm3 for prostates with a mean measured volume of 30 cm3 or less. These numbers increased to 6.84 cm3 and 3.99 cm3, respectively, when measurements were performed by one experienced and one less experienced observer (110 prostate volumes measured in this case). In low-dose rate (LDR) permanent implant brachytherapy, for ease of planning and more robust seed implantation, some centers prefer contours that are smooth and symmetric with respect to the medial line (6) in the transverse plane. These two requirements are difficult to satisfy manually, whereas an automatic segmentation method, in addition to producing a smooth and symmetric volume, can reduce the variability of the contours related to the observer bias and random factors. Additionally, the time required for performing segmentation can be greatly reduced, and thus can be adapted for subsequent intraoperative planning. Various ultrasound-based segmentation methods have been proposed in the literature. Methods using higher-level knowledge, such as using 3D geometric shapes, in addition to lower-level image information, such as texture and edges, have been more successful compared with pure image-based segmentation methods.