Shift during Neurosurgery

Abstract. Intraoperative imaging is used more frequently today since this en-ables the surgeon to localize structures inside the brain more accurately and helps to detect shape changes. In order to combine informations derived at different time points, we describe a nonrigid registration algorithm that aligns MR scans of the brain. Based on a set of feature points, an initial sparse estimate of the displacement field is found by optimizing a local cost function. A linear elastic model is then used to infer the volumetric deformation across the image. Inhomogeneous elasticity parameters are generated using empirically observed variability of the brain from a dataset of 154 young adults. Initial results are generated on intraoperative image sequences showing brain shift.