Yixun Liu1, Zhijian Song. 1. Digital Medical Research Center of Fudan University, Shanghai 200032, People's Republic of China.
Abstract
BACKGROUND: Brain deformation plays an important role in causing inaccuracy in image-guided neurosurgery. Three types of approaches have been proposed to solve this problem: intra-operative imaging, deformation atlas and non-rigid registration. By comparing these approaches, we here show that the non-rigid registration approach, based on a linear elastic model, may be the most feasible method during clinical application. METHODS: Based on the non-rigid registration model, we designed a framework used to correct the brain deformation. A laser range scanner (LRS) was introduced into this framework to obtain the intra-operative brain surface. Using this device, we designed a novel surface-tracking algorithm, which includes space transformation (rigid registration) and surface moving. We first transformed the point set from LRS space into image space by a series of transformations, then simulated the movement of the brain surface using a thin-plate spline. RESULTS: We tested the framework using pigs. In these experiments, we segmented and meshed the pig's brain and transformed the initial surface (from a MRI scan) and deformed surface (from LRS) into the same coordinate system, using rigid registration. Using this method, the surfaces of pigs' brains were tracked accurately and the internal brain deformation was estimated. The pre-operative images can be corrected accordingly. CONCLUSIONS: Our animal experiments indicate that this framework can effectively capture the surface deformation and hence estimate the internal deformation of the brain. (c) 2008 John Wiley & Sons, Ltd.
BACKGROUND: Brain deformation plays an important role in causing inaccuracy in image-guided neurosurgery. Three types of approaches have been proposed to solve this problem: intra-operative imaging, deformation atlas and non-rigid registration. By comparing these approaches, we here show that the non-rigid registration approach, based on a linear elastic model, may be the most feasible method during clinical application. METHODS: Based on the non-rigid registration model, we designed a framework used to correct the brain deformation. A laser range scanner (LRS) was introduced into this framework to obtain the intra-operative brain surface. Using this device, we designed a novel surface-tracking algorithm, which includes space transformation (rigid registration) and surface moving. We first transformed the point set from LRS space into image space by a series of transformations, then simulated the movement of the brain surface using a thin-plate spline. RESULTS: We tested the framework using pigs. In these experiments, we segmented and meshed the pig's brain and transformed the initial surface (from a MRI scan) and deformed surface (from LRS) into the same coordinate system, using rigid registration. Using this method, the surfaces of pigs' brains were tracked accurately and the internal brain deformation was estimated. The pre-operative images can be corrected accordingly. CONCLUSIONS: Our animal experiments indicate that this framework can effectively capture the surface deformation and hence estimate the internal deformation of the brain. (c) 2008 John Wiley & Sons, Ltd.