Publication Details
Yixun Liu, Chengjun Yao, Fotios Drakopoulos, Jinsong Wu, Liangfu Zhou and Nikos Chrisochoides.
Published in Medical Physics, Publisher American Association of Physicists in Medicine, doi: 10.1118/1.4893754, Volume 41, No. 10, August, 2014
Abstract
Purpose: This paper presents a non-rigid registration method to align preoperative MRI (preMRI) 35 with intra-operative MRI (iMRI) to compensate for brain deformation during tumor resection. This method extends traditional point-based non-rigid registration in two aspects: 1) allow the input data to be incomplete, and 2) simulate the underlying deformation with a heterogeneous biomechanical model. Methods: The method formulates the registration as a three- variable (point correspondence, 40 deformation field, and resection region) functional minimization problem, in which point correspondence is represented by a fuzzy assign matrix; Deformation field is represented by a piece-wise linear function regularized by the strain energy of a heterogeneous biomechanical model; and resection region is represented by a maximal simply connected tetrahedral mesh. A Nested Expectation and Maximization framework is developed to simultaneously resolve these three 45 variables. Results: To evaluate this method, we conducted experiments on both synthetic data and clinical MRI data. The synthetic experiment confirmed our hypothesis that the removal of additional elements from the biomechanical model can improve the accuracy of the registration. The clinical MRI experiments on 25 patients showed the proposed method outperforms the ITK implementation of a physics-based 50 non-rigid registration (PBNRR) method. The proposed method improves the accuracy by 2.88 mm on average when the error is measured by a robust Hausdorff distance metric on Canny edge points, and improves the accuracy by 1.56 mm on average when the error is measured by 6 anatomical points. Conclusions: The proposed method can effectively correct brain deformation induced by tumor resection.
