Modeling Brain Retraction In Neuronavigation

Most of brain tumors locate in deep brain tissue. During neurosurgery, in order to expose the targeted tumor, doctors may need retractors to separate brain tissue. This may cause brain retraction which is a type of brain shift. This brain shift leads to great distortion to Image Guided Neurosurgery System (IGNS) which is constructed basing on preoperative images of the patients. Therefore this guiding systems fall into nothing but a series of useless images.We, firstly, listed three types of main methods modeling brain retraction with two of them referring to Finite Element Method (FEM), one referring to eXtended Finite Element Method (XFEM). And then, we applied all three methods into a single type of cube building on the same linear elastic model, which aimed to quantitatively evaluate three methods. Since brain retraction exits topological discontinuity, after the deformed meshes and deformed images were collected, we found that FEM could hardly model this discontinuity accurately. Instead, XFEM could handle this. Based on this exciting discovery, we employed a series of preoperative CT images from a human brain-like phantom and used XFEM to deform a linear elastic model with the help of Boundary Conditions (BCs) acquired by intraoperative CT scanning. This experiment helps us evaluate XFEM more objectively. XFEM did well, however the methods acquiring BCs was quite troublesome. We brought Laser Range Scanners into application of brain retraction modeling. Accompanied by brain shift surface tracking algorithm, this innovative framework could rapidly collected surface point clouds of deformed brain and transformed these point sets into useful BCs. At last, XFEM used these BCs to quickly warp to the preoperative images. This new framework then was applied to one brain phantom, one swine brain in vitro and one live swine.The forecast error of the framework, which designed to validate XFEM and applying BCs of one brain phantom acquired by iCT, varied between0.30mm and0.50mm (mean0.40mm), and its correction accuracy increased from83.1%to87.5%(mean85.9%). The forecast error of the framework, which designed to validate XFEM and applying BCs of one brain phantom acquired innovatively by one LRS, varied between0mm and1.73mm (mean1.19mm), and its correction accuracy increased from51.4%to100%(mean81.4%). Besides, the modified Hausdorff distances extracted from the contours of preoperative images the deformed images and intraoperative images decreased from1.10mm to0.76mm. The forecast error of the same framework equipped with XFEM calculation and LRS intraoperative scanning on the swine brain in vitro varied between0.00mm and2.00mm (mean0.82mm), its correction accuracy ranged between41.9%and100%(mean83.1%). At same time, when this framework took into use on the live swine, the forecast error varied from0.00mm to0.96mm (mean0.40mm), the correction accuracy ranged between66.9%and94.8%(mean80.5%).Brain retraction causes topological discontinuity of the brain tissue. Compared to FEM, XFEM could better model this discontinuity. Via building one linear elastic model based on preoperative images and deform this model by XFEM, the image quality which finally acquired was much better, so that these images could continuously guide doctors to finish the surgery. Since the BCs acquired by iCT were limited, LRS equipped with a brain shift surface tracking algorithm could extract BCs more rapidly and easily. With the model calculation of XFEM which equipped with LRS scanning, our framework seems more convincing, and is much easier to put into clinical applications.