| During neurosurgery, nonrigid brain deformation, referred to as brain shift, prevents preoperatively acquired images from accurately depicting the intraoperative brain. Image-guided surgical navigation systems, therefore, must account for this brain shift in order to provide accurate surgical guidance. However, the origins and complexity of this type of deformation prevent it from being entirely predicted preoperatively. Additionally, though volumetric images can be acquired at the time of intervention, this type of intraoperative imaging is either expensive, invasive or time intensive. A solution that overcomes these issues consists of warping preoperative images to reflect the intraoperative brain using sparse intraoperative information. One such source of intraoperative information, the exposed cortical surface, can be tracked optically, for example, using stereo vision. Unfortunately, however, these systems are often plagued with calibration error, which can corrupt the surface deformation estimation.; In order to separate the effects of camera calibration and surface deformation, a framework is needed which can solve for disparate and often competing variables. In this work, game theory', which was developed specifically to handle decision making in this type of competitive environment, has been applied to the problem of cortical surface tracking and used to infer information about the physical processes of calibration and brain deformation. The specific application of this work is neocortical epilepsy, in which information about the surface deformation is the most critical. However, it is also shown that this type of surface deformation estimation can be extended to the volume through the use of a biomechanical model.; As with any method that will be used in vivo, it was imperative to validate the algorithm results before patient application. For this purpose, a realistic brain phantom was constructed, which could simulate the brain shift experienced during surgery. The algorithms were tested both in simulation and using the realistic phantom. The result was a reliable intraoperative tracking method, which was tested on eight in vivo data sets. This ultimate goal of this project is to provide neurosurgeons with accurate surgical guidance, allowing better detection of pathologic tissue and decreased neurosurgical complications. |
