Inferring brain deformation during open neurosurgery using CBCT angiography

During neurosurgery the brain can deform, which is known as brain shift. The deformation can be as large as 20 mm and limits the accuracy of image-guided surgery. In this work we use pre- and intra-operative CBCT angiography to infer this deformation. We first segment cortex and vessels in both images and then perform surface registration between the pre- and intra-operative segmentations. The resulting deformation field is assumed to correspond to the brain deformation and can be used to update other pre-operative images. Since the intra-operative image quality is deteriorated by streak artifacts originating from the metallic stereotactic frame, our approach explores robust segmentation and registration methods that reduce the influence of these artifacts. Our method is tested on a clinical data set of 8 patients who underwent neurosurgery. We also compare with a recently published method and rigid registration as a baseline method, showing the improved accuracy of our approach.