GammaPlan MR to CT Image Registration Errors: Implications for Extend and Preplanned Treatments
A Sudhyadhom*, A Perez-Andujar, L Ma, I Barani, UCSF Comprehensive Cancer Center, San Francisco, CASU-E-J-89 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: To compare the spatial errors in coordinate localization created by MR-based stereotaxia in GammaPlan for two different methods: direct stereotactic MR localization and MR co-registration with a stereotactic CT.
Methods: Seven patients underwent Gamma Knife SRS and were scanned with our clinical stereotactic MR and stereotactic CT protocols. In GammaPlan, the stereotactic coordinate system for each MR and CT was defined using the fiducials of the localizer. Separately, each MR was co-registered (by the patient's anatomy) to the stereotactic CT using GammaPlan's automated co-registration algorithm. A rigid transformation relationship was determined between the direct stereotactic MR (stereoMR) and the MR co-registered to the stereotactic CT (coregMR). Spatial errors in coordinate definition between these two methods were calculated at the center of stereotactic space and at the centroid of the target. A total of seventeen MR scans were analyzed by this method including both axial and coronal acquisitions.
Results: Mean errors in fiducial definition were found to be 0.6±0.3mm and 0.7±0.3mm for CT and MR, respectively. At the center of stereotactic space, the mean magnitude error in coordinate localization between stereoMR and coregMR was found to be 1.6±0.6mm. At the centroid of the target, the mean magnitude error was found to be 1.5±0.7mm. These results were statistically significant compared to the errors in fiducial definition (p<0.002 for both). No statistically significant systematic errors along any axes were found in comparing stereoMR and coregMR coordinate localizations.
Conclusion: Statistically significant differences were found in coordinate localizations between the two methods. These errors may primarily be due to sub-optimal co-registrations when using GammaPlan's automated registration algorithm. The magnitude of these errors may be clinically significant as they are on average greater than 1.5mm. Errors in co-registration are important for preplanned SRS and Extend SRT treatments where direct stereotactic MR localization is not utilized.