Automated brachytherapy seed localization using intensity-weighted feature extraction techniques
Brachytherapy has proven to be an effective treatment of various forms of cancer, whereby radioactive material is inserted directly into the body to maximize dosage to malignant tumors while preserving healthy tissue. In order to validate the pre-operative dosimetric model, a post-implant evaluation procedure is needed to ensure that the locations of the implanted seeds are consistent with the planning stage. Physicians require an automated algorithm for seed localization to expedite the post-implant evaluation process and reduce human error. Previously developed algorithms have performed binary transforms on images before attempting to localize seeds. Furthermore, traditional approaches based upon three-dimensional seed shape parameterization and matching require high resolution imaging. We propose a new algorithm for automatic seed localization for full three-dimensional data sets that utilizes voxel intensity for both seed centroid computation and estimation of angular seed orientation that is expected to produce reliable results even for low-resolution images. © 2007 IEEE.