Nonrigid brain registration: synthesizing full volume deformation fields from model basis solutions constrained by partial volume intraoperative data

Abstract

During image-guided neurosurgery, maintaining accurate registration of the patient with the preoperative image volume is essential to any navigational system. Since the patient’s brain shifts during many OR procedures, we have developed a physically based deformation model to update images concurrent with surgery in order to achieve nonrigid registration between the brain and the preoperative scans. In this paper, we introduce a strategy for integrating sparse displacement data acquired during surgery with the computational model using an efficient and accurate approach. The complex boundary conditions that exist during surgery are estimated from sparse data through a synthesis of simpler precomputed sets of model basis solutions. These basis solutions are weighted in accordance with a minimization procedure that reduces the error between the observed and computed displacement fields. This method appears to be a promising technique for increasing the speed and accuracy of the computational estimate, thus making intraoperative updates more efficient. Furthermore, it has the advantage of incorporating intraoperatively acquired measurements of true displacements into the model to ensure a more accurate estimation of tissue motion. Results from in vivo pig brain experiments involving multiple retractions show that full volume deformation fields can be constructed throughout a series of retraction events from a single set of basis solutions with equal or increased accuracy.