Publication Details
Andriy Fedorov, Andriy Kot, Yixun Liu, Olivier Clatz, Peter Black, Alexandra Golby, Ron Kikinis and Nikos Chrisochoides.
Published in Book on DDDAS- Dynamic Data Driven Applications Systems, Publisher Springer Verlag, 2010
Abstract
Current neurosurgical procedures utilizemedical images of variousmodalities to enable precise location of tumor and critical brain structures for the purposes of planning accurate brain tumor resection. The practical difficulty of using preoperative images during the surgery is caused by the intra- operative deformation of the brain tissue (brain shift), which introduces discrepancies with respect to the pre-operative configuration. Intra-operative imaging allows tracking of such deformations, but cannot fully substitute for pre-operative data. Non-Rigid Registration (NRR) is a complex time-consuming image processing operation that allows the adjustment of the pre-operative image data to account for intra-operative brain shift. We review computational aspects of a specific method for registering brain MRI to enable its evaluation during image-guided neurosurgery, and consider different strategies for parallelizing this NRR method. We show that the implementation we develop not only allows the delivery of NRR results within the clinical time constraints improving NRR speed, but also provides the potential of improving the accuracy of registration by utilizing distributed Grid resources for distributed search of optimum parameters for the NRR method. In this context, we describe a concept of a dynamic data driven environment for highly distributed non-rigid registration calculations.We present initial results of using national cyberinfrastructure as a platform for such environment, and outline the major challenges in integrating it with the operating rooms of the future.
