Title Page Title Non-­‐rigid Registration for Image Guided Surgery Character 47

Objective: Endoscopic skull base or cranial procedures can cause brain shift. Currently useful intra-­‐operative soft tissue information is obtained via intraoperative MRI (iMRI). Intraoperative CT (iCT) has been introduced but soft tissue information is limited. Therefore our objective is to determine if non-­‐rigid registration (NRR) between iCT images and pMRI images is possible for intraoperative soft tissue movement. Methods: Patients with skull base pathology with compression of soft tissue were recruited. pCT and pMRI images were obtained. Segmentation of the radiographic images was performed to create a tetrahedral volume mesh. Anatomic features were identified for " block matching ". iCT images were acquired and rigid registration (RR) for iCT to pCT and pMRI were performed. Block matching between the iCT and the pCT was executed solving for the dense deformation field. NRR was achieved with the pMRI according to the dense deformation field. Matching and registration error were calculated and compared between RR and NRR. and dense deformation field estimation was successfully performed. The NRR accuracy improved on an average of 4.14 times compared to RR. The registration accuracy for acquired deformation improved by 19% for NRR over that of RR. This is the first study evaluating the NRR of iCT with pMRI. NRR is possible between the pCT and iMRI. Successful application of this technique would allow iCT to become a pseudo-­‐iMRI at a fraction of the cost, however clinical application has not been determined.