A Template Based Approach for Automatic Seed Detection in Post-implant CT Images for Prostate Brachytherapy

3D reconstruction of implanted seeds is an essential step for post-operative dosimetry in prostate brachytherapy. However, because of the slice thickness and artifacts produced by seeds, large areas will be appeared in computed tomography (CT) images representing many seeds. We propose a new algorithm for the automatic localization and reconstruction of implanted seeds with main focus on localizing connected seeds using a set of binary templates with respect to the shape of artifacts of individual and connected seeds. The results compared with those of fluoroscopic images demonstrate that our algorithm can achieve above 97% detection rate with less than 0.5 mm localization error and reliable orientation accuracy.