Evaluation of accuracy of patient-specific non-linear biomechanical models for predicting intra-operative brain shift

[Truncated abstract] Malignant glioma is the most common primary brain tumour in adults. It generally presents with epilepsy, cognitive change, headache, dysphasia, or progressive hemiparesis. Diagnosis is usually achieved by appropriate imaging studies followed by biopsy or neurosurgical resection. Near-total surgical removal is desirable for delayed malignant progression, decreased risk of seizures and prolonged survival. This is difficult to achieve due to the uncertainty in visual distinction of gliomatous tissue from adjacent healthy brain tissue. More complete tumour removal can be achieved through image-guided neurosurgery that uses intra-operative magnetic resonance images (MRIs) for improved visualization. The efficiency of intra-operative visualization and monitoring can be significantly improved by fusing high resolution pre-operative imaging data with the intra-operative configuration of the patient’s brain. This can be achieved by updating the pre-operative image to the current intra-operative configuration of the brain through registration. During neurosurgery, however, the brain deforms, up to 20 mm in some cases. This deformation or brain shift is due to several factors such as- loss of cerebrospinal fluid (CSF), gravity, administered drugs, tissue resection and removal, intracranial pressure, etc. and should be taken into account while registering images. While intra-operative imaging is the most straightforward way to capture brain deformation during surgery intra-operative MRI scanners are very expensive and often cumbersome. Hardware limitations of these scanners make it unfeasible to achieve frequent whole brain imaging during surgery. As an alternative the pre-operative MRI can be updated to the current configuration of the operating room. The contemporary way to update the pre-operative image is to non-rigidly register it with the intra-operative image. However, this approach still requires frequent acquisition of the intra-operative images. An alternative approach is to acquire very rapid sparse intra-operative data and predict the deformation for the whole brain. The Intelligent Systems for Medicine Laboratory (ISML) at UWA has developed a suite of algorithms for real time prediction of the brain deformation from sparse intra-operative data. The objective of this work is to prove that the new specialised non-linear finite element algorithms developed by ISML should enable accurate estimation of brain shift. Rigid registration is a technology currently available to patients. The accuracy of registration results obtained from two algorithms are compared in this study- (1) biomechanics-based Total Lagrangian Explicit Dynamics (TLED) algorithm that uses only the intra-operative position of the exposed surface of the brain; and (2) Rigid registration algorithm as implemented in 3D Slicer that uses intraoperative image as the target image. Results of 33 neurosurgery cases are compared. In order to estimate registration error a quantitative measure of misalignment between two images is required...