ABSTRACT
This study was performed to develop and evaluate a semi-automatic seed localization algorithm from magnetic resonance (MR) images for interstitial prostate brachytherapy. The computerized tomography (CT) and MR images (3 mm-slice thickness) of six patients who had received real-time MR imaging-guided interstitial prostate brachytherapy were obtained. An automatic seed localization method was performed on CT images to obtain seed coordinates, and an algorithm for seed localization from MR images of the prostate was developed and tested. The resultant seed distributions from MR images were then compared to CT-derived distribution by matching the same seeds and calculating percent volume receiving 100% of the prescribed dose and the extent of the volume in 3-dimensions. The semiautomatic seed localization method made it possible to extract more than 90% of the seeds with either less than 8% of noises or 3% of missing seeds. The mean volume difference obtained from CT and MR receiving 100% of the prescribed dose was less than 3%. The maximum extent of the volume receiving the prescribed dose were 0.3, 0.6, and 0.2 cm in x, y, and z directions, respectively. These results indicate that the algorithm is very useful in identifying seeds from MR image for post-implant dosimety.