Publication Details
Fotios Drakopoulos, Chengjun Yao, Yixun Liu and Nikos Chrisochoides.
Published in MICCAI Computational Biomechanics for Medicine CBM XI Workshop, Athens, Greece, October, 2016
Abstract
Interventional MRI (iMRI) has proven to be an effective tool for the visualization of brain deformation and for the optimization of the maximal safe volumetric tumor resection during Image-Guided Neurosurgery. Earlier we proposed two adaptive non-rigid registration methods between pre-operative and intra-operative MRI based on: (i) Nested Expectation Maximization (NEM) which implicitly compensates for tissue removal, and (ii) Geometric-based which explicitly adapts the mesh to compensate for the changes in the geometry of the brain. In this paper, we assess the accuracy of these methods and compare them with two widely used registration schemes: ITK’s rigid registration, and ITK’s Physics Based Non-Rigid Registration (PBNRR). The evaluation is based on registration error, for the brain deformation induced by cerebral glioma resection, and it utilizes three metrics for the error: (i) 100% Hausdorff Distance, (ii) error at specific anatomical points identified by a neurosurgeon, and (iii) visual inspection by a neurosurgeon. We conduct a retrospective study on ten patients with a malignant glioma. The evaluation shows that the geometric adaptive approach achieves the most accurate alignments compared to ITK’s PBNRR and the Nested Expectation Maximization. It significantly reduces the alignment error due to rigid registration commonly used by commercial neuronavigators, and completes a volumetric alignment, on average, in about 2.3 minutes on a 12-core Linux workstation, satisfying the time constraints imposed by neurosurgery.
