Combined feature/intensity-based brain shift compensation using stereo guidance

During neurosurgery, soft tissue deformation produces nonrigid brain motion. Biomechanical models are often used in conjunction with image-derived information to infer volumetric brain displacements and compensate for this deformation. Proper use of these compensation systems depends on incorporating appropriate model parameters, balancing the model/data tradeoff and, importantly, on the accuracy of the image-derived information used with the model. The goal of this work is to improve cortical surface tracking accuracy using intraoperative stereo camera images. We use image-derived cortical surface displacement to drive our model. This method takes advantage of both stereo image intensities and segmented cortical features to detect surface motion within a Bayesian framework. To quantify accuracy, the algorithm is tested on both simulated and real surfaces.