The use of mutual information in registration of CT and MRI datasets post permanent implant.

PURPOSE To determine the feasibility of registration of MRI and CT datasets post permanent prostate implant by the use of mutual information. METHODS AND MATERIALS Five patients who underwent permanent (125)I implant for prostate carcinoma were studied. Two weeks postimplant an axial CT, T2-weighted-axial, sagittal and coronal MRI, and T1-fat-saturation MRI scans were obtained. Registrations of MRI to CT and MRI to MRI datasets were performed by mutual information, an automated process of data registration matching all information in specified dataset regions of interest. Registration quality was evaluated by visual inspection, agreement with seed- to-seed registration, and histogram analysis. RESULTS Rapid registration (<30 minutes) of CT and MRI datasets can be accomplished through the use of mutual information. All methods of registration evaluation confirmed excellent registration quality. Although D90 and V100 for the prostate were comparable between MRI- and CT-based dosimetry, dose to critical structures/microenvironments (anterior base, posterior base, bladder outlet, lower sphincter, bulbar urethra) defined on MRI varied widely. CONCLUSIONS Efficient and accurate registration of MRI and CT datasets following prostate implant is possible, and improves the accuracy of postimplant dosimetry by superior definition of the prostate. Definition of critical microenvironments and adjacent structures will improve dose and toxicity correlation and ultimately improve planning strategies.