Computing Reaction Forces on Surgical Tools for Robotic Neurosurgery and Surgical Simulation Adam Wittek 1 , Jerome Laporte 1 , Karol Miller 1 1 Intelligent Systems for Medicine Laboratory School of Mechanical Engineering, The University of Western Australia 35 Stirling Highway, Crawley/Perth WA 6009, AUSTRALIA kmiller@mech.uwa.edu.au http://www.mech.uwa.edu.au/ISML Ron Kikinis 2 , Simon K. Warfield 3 2 Surgical Planning Laboratory, 3 Computational Radiology Laboratory Radiology, Brigham and Women’s Hospital and Harvard Medical School 75 Francis Street, Boston, MA02115, USA http://splweb.bwh.harvard.edu:8000/ Abstract The objective of our research is to create a system computing brain deformations. In this paper we concentrate on assessing the feasibility of using non-linear finite element computation for patient-specific simulations. As an example we use computation of reaction forces on surgical tools, with application in e.g. surgical robot control system, virtual reality operation planners, etc. We specifically address issues related to creating geometrically and mechanically precise representations of the brain. The method comprises of the following steps: 1) development of a “generic” brain mesh; 2) conversion of the generic brain mesh to patient- specific brain mesh; 3) selection of the appropriate mathematical model of the brain biomechanics; and 4) development of an efficient computational scheme. As an illustration of the presented concepts we provide an example of 3D meshing, and calculation of reaction force acting on a surgical tool using a single-phase mathematical model solved using an explicit, non-linear finite element procedure. 1 Introduction The advantages of surgical robots and manipulators are well recognised in the clinical and technical community. Precision, accuracy and the potential for telesurgery are the key motivators for applying advanced robots in surgery [Chinzei et al., 1999; Chinzei and Miller, 2001]. Surgical robots require trajectory planning, which in practice relies upon the preoperative images. However, if the organ deforms, the trajectory needs to be updated during the procedure. Although nuclear magnetic resonance images (NMRI) can provide rich information of tissue deformation [Grimson et al., 1999; Kaus et al., 1999], currently tens of seconds are required to produce and analyse a new set of images. One possible method of dealing with these delays can be predicting the organ deformation by means of computer simulation. When the organ deformation can be predicted and the tissue mechanical properties are known, the interaction force between the end-effector (i.e. surgical tool) and organ surface during surgery can also be computed. Thus, simulation can be used to predict reaction forces acting on surgical tools (equal to forces with which the tool acts on the organ) as well as to provide realistic force and tactile feedback for virtual reality surgical simulators. To achieve a realistic, clinically acceptable computer simulation of human body organ deformation during surgical procedure, one requires an adequate model, capturing the intrinsic physical properties of the organ considered, intervention performed, and surgical accessories used. Neurosurgery is particularly demanding, as the brain is arguably the most complicated object in the known universe. Modelling of physical properties of the brain is still an uncovered area pioneered by a few only [Miller and Chinzei, 1995; Bilston et al., 1997; Paulsen et al., 1999; Prange and Margulies, 2002]. The models used in surgery simulation should contain detailed anatomical (geometrical) information. Such information can be provided by suitable anatomical atlases [Talairach and Tournoux, 1988; Visible Human, 1995; Nowinski, 2003]. However, simulation of surgical procedures typically requires patient-specific geometric information. Therefore, the information provided by the atlas has to be individualised through an appropriate scaling process to fit to a particular patient, referred to as registration, see e.g. [Warfield, 1999; Ferrant, 1999; Ferrant, 2000; Ruiz-Alzola, 2000]. In this process the medical radiographic images of the patient brain are utilised. Next, a computational grid has to be created on the domain of interest. In most practical cases this amounts to producing a finite element mesh. 1