Publication Details
Yixun Liu, Chengjun Yao, Liangfu Zhou and Nikos Chrisochoides.
Published in IEEE International Symposium on Biomedical Imaging: From Nano to Macro, pages 1217 -- 1220, April, 2010
Abstract
This paper presents a novel feature point based non-rigid registration of preoperative MRI with resected intra-operative MRI (iMRI) to compensate for brain shift during tumor resection. The registration is formulated as a three variables (Correspondence, Deformation Field and Resection Region) functional minimization problem. We solve this problem by means of a nested Expectation and Maximization (EM) framework where: (1) the inner EM loop computes the Correspondence and Deformation Field by iteratively rejecting point outliers and (2) the outer EM loop computes the Resection Region by iteratively rejecting resected elements. Our preliminary data from both synthetic and real brain MRI show the effectiveness of this method to handle tumor resection. The results of the registration in the vicinity of the tumor resection is on average, 16 times more accurate than the results from rigid registration.
