Automatic localization of implanted seeds from post-implant CT images

An automatic localization method of implanted seeds from a series of post-implant computed tomography (CT) images is described in this paper. Post-implant CT studies were obtained for patients who underwent prostate brachytherapy. Bright areas were segmented using binary thresholding in each CT slice, and geometrical information on these areas was collected. Large areas (possibly containing two connected seeds) were split into smaller ones by geometry-based filtering in each slice. The area connectivity along the longitudinal direction was analysed using a geometry-based connection search algorithm executed on every area slice by slice, so that the connected areas were combined into one object. The weighted centroid of each object was taken as the seed position. This method was tested on a seed-containing prostate phantom as well as using CT studies from patients. Statistical analysis demonstrates that it can achieve above 99% detection rate with reliable localization accuracy and high speed. It is reliable and convenient for localizing implanted seeds on CT and can be used to assist post-implant dosimetry for prostate brachytherapy.