Po‐Poster ‐ 26: Investigation of normalized mutual information for co‐registration of CT — MR images of permanent prostate implants

Abstract

Post‐implant dosimetric evaluation is the standard contemporary method for assessing permanent prostate implant quality and determining the dose received by the prostate and organs at risk over the course of treatment. In current practice evaluation is performed using a CT image set typically acquired one month after the implant. However, due to poor visualization of prostate contours on the CT images, the evaluation can be difficult and is only approximate. MRI allows a better appreciation of prostate contours, but the visibility of the ensemble of the seeds is insufficient to allow its use as the only means of evaluating the treatment. Image registration of CT and MR image sets can combine the benefits of both imaging modalities and provide the means for sufficiently accurate dosimetric analysis. Mutual information is an efficient registration technique that can be used to fuse selected volumes of interest such as the prostate and neighbouring tissues. In our study we investigate optimization of registration search parameters, such as threshold level, translation and rotation angle ranges, and volume boundaries for fusing CT — MR images of a tissue equivalent prostate phantom implanted with inactive seeds, simulated image data, and clinical data sets. The results obtained so far are encouraging and suggest that mutual information‐based registration may eventually enable automatic fusion of clinical CT and MR images for prostate implant post‐dosimetry.