ECE Graduate Seminar

You are cordially invited to attend the following seminar:   Department of Electrical & Computer Engineering   Strategies for Achieving Patient-specific High-validity Simulation and Descriptive Navigation for Neurosurgery   by Dr. Michel Audette Department of Modeling, Simulation and Visualization Old Dominion University   Friday, April 25, 2014 3:00 p.m. KH 224 Host:  Dr. Oscar González   This presentation describes techniques for achieving patient-specific high-validity simulation and descriptive planning of neurosurgery. Broadly speaking, my research is dedicated to the long-term development of a broadly usable patient-specific, high-validity neurosurgery simulator, which in turn entails modeling both the anatomy and therapy. Moreover, any innovation in anatomical modeling elaborated for simulation purposes can be applied to planning for image-guided neurosurgery (IGNS). Furthermore, research in both simulation and planning of neurosurgery can also be leveraged to facilitate surgical robotics. I will describe my modus operandi in research, which also has implications for teaching, which aggressively leverages open-source software, such as Simulation Open Framework Architecture (SOFA) for simulation and Insight Segmentation and Registration Toolkit (ITK) for medical image analysis, as well as contributed source code; this approach also potentiates another characteristic of my work, mentoring undergraduate and high-school research.  I will discuss computer-assisted neurosurgery, which subsumes IGNS, simulation, and robotics, as well as of the importance of developing clinical requirements for these technologies. It is essential to develop IGNS, simulation and robotics that fulfill practical, well-defined clinical requirements; I will allude to methodologies planned for producing these requirements. These methodologies will ultimately lead to the development of an ontological representation of neurosurgery where each procedure is expressed in terms of broad stages, organized according to two parameters: i) the neurosurgical approach (pterional, transnasal, and so on), with implications in terms of which anatomy to emphasize, and ii) the nature of the pathology - tumor type, arteriovenous malformation (AVM), deep-brain stimulation (DBS) target, and so on, with implications for the choice of therapy. In simulation, our objective is high predictive validity, which is a representation of future real-world performance, as opposed to “lesser” types of validity. In surgical navigation and robotic surgery, I advocate the most descriptive anatomical modeling possible, fulfilling practical requirements of neurosurgeons, featuring critical tissues at risk as well as an efficient means of estimating intra-surgical brain shift. In the area of anatomical modeling, our techniques address the main computational stages that produce models of the brain and spine from medical images: segmentation, which maps intensities to tissues, and 3D meshing, which decomposes tissues into simple shapes such as tetrahedra for simulation as well as accurate, brain-shift compensated IGNS, which also applies to robotic surgery. In general, I am a proponent of segmentation that is based on a digital functional atlas as well as active surface and contour models. Furthermore, rigorous clinical requirements dictate that we exercise care with the type of image data that we employ, which suggests leveraging recent development in MRI pulse sequences. MRI not only provides morphological information with T1, T2 or proton density weighting, but can provide multi-contrast, tractographic, angiographic, metabolic as well as elastographic imaging. I will present on-going work of my group in the development and application of digital atlas development and active surface and contour models. One of our main tools is the active simplex model, which is a physically based active surface mesh model, where each vertex is subject to Newtonian mechanics and is linked to 3 neighboring vertices by edges (in the 2-simplex case), which in turn leads to a triangulated surface by geometric duality. As a result, the final simplex-identified anatomical boundary can provide triangulated surface boundaries of the anatomy by duality,  which are then used as input boundaries to a subsequent tetrahedral meshing stage. Our variant of the simplex, contributed by a collaborator, can reconcile multiple anatomical boundaries and prevent spatial overlap between them, in contrast to merely a single, inside-outside surface boundary; it also offers precise control over surface mesh size, while producing simplex faces and dual triangular faces of high quality. We view both multi-surface representation and resolution control as essential objectives in developing interactive surgery simulation, in conjunction with a multi-grid approach to interactive biomechanics, which will build on SOFA software classes developed collaboratively. In particular, we want to be able to represent the brain at a coarse resolution with a few thousand tetrahedra, conducive to an interactive  tissue response, a medium-resolution inner volume along the surgical path based on the neurosurgical approach, as well as a fine-resolution volume of the tumor, AVM or DBS target and critical tissues nearby.  I will present adaptations of the multi-surface simplex underway by my graduate students: we are developing minimally supervised model-based approaches for segmenting, for IGSN and simulation: i) subcortical structures for DBS, ii) the spine for discectomy procedures, and iii) the skull base anatomy, with an emphasis on cranial nerves.  The first project is a multi-surface mesh representation of a digital deep brain atlas, where multi-material contouring will lead to a multi-surface functional brain mesh with shared boundaries. In conjunction with a MRI-compatible DBS robot, whose development is underway, this multi-surface deep-brain mesh atlas will exploit image forces, characteristic of our simplex model, that enable non-rigid registration with both preoperative and intraoperative patient MRI data, thereby enabling descriptive planning and accurate navigation, brain shift-compensated in seconds, through a lightweight surface mesh representation of subcortical targets such as the subthalamic nucleus. The second is the development of a multi-surface spine model, where we will combine multi-surface modeling with population statistics of shape, while turning off these statistics in pathological areas.  The third is the development of a model of the skull base anatomy, emphasizing cranial nerves, where we are replacing the usual 2-simplex framework, appropriate for surfaces, with a 1-simplex appropriate for 3D curves; we will adapt the simplex’ contact handling to make it possible to prevent overlap with cerebrovascular structures nearby. This project will also build on an undergraduate research project, also underway, that will transpose a printed atlas of the brainstem into 3D digital format, making explicit the nerves’ attachment points.  I will also describe some of the objectives planned in the area of therapy simulation, which for interactive simulation will build on the SOFA platform. This area of research features on-going work of a student in our department, to which I am contributing, which employs a SOFA-based meshless approach to interactive cutting simulation. A student in my group has begun building on this effort, as well as prior work of mine on 7 degree-of-freedom haptics-based bite modeling, to towards the simulation of the bite of a Blakesley forceps, which will be used in our discectomy simulation. I will describe other projects planned for the near future that relate to therapy simulation, which includes predictive modeling of the ultrasonic surgical aspirator, the main tool used for neurosurgical tumor resection, and plans for the patient-specific, white-matter tract-aware simulation of deep-brain stimulation.   Bio Dr. Michel Albert Audette is an Assistant Professor in the Department of Modeling, Simulation and Visualization Engineering (MSVE) at Old Dominion University. He joined ODU in August 2011. Prior to joining ODU he worked at Kitware, Inc., Chapel Hill, NC working on image analysis and medical simulation. His academic training includes:   McGill University, Montreal Electrical Engineering B.Eng., 1986 Ecole Polytechnique, Montreal Electrical Engineering M. Eng., 1992 McGill University, Montreal Biomedical Engineering Ph.D., 2002 AIST, Tsukuba, Japan Medical Simulation 02/2001-12/2005 ICCAS, Leipzig, Germany Medical Simulation 01/2006-11/2008   At ODU he has developed courses in medical image analysis and interactive simulation. He is an active collaborator of a Simulation Open Framework Architecture (SOFA) interactive simulation textbook and curriculum. He is also mentoring undergraduate and high school students.

Posted By: Linda Marshall
Date: Wed Apr 23 09:54:43 EDT 2014