Virtual paracentesis simulation snapshots: (a) solid and (b) alpha-transparency rendering. 

Context in source publication

Context 1

... Figure 2 shows two screen snapshots of the virtual paracentesis simulation, including 3D renderings of the human skin in solid and alpha-transparency mode (with visible bony structures, lungs, and part of the vascular tree involved in the considered paracentesis operation). In the current simulator configuration, the system continuously monitors the trajectory of the needle, and its intersection with the surface of the skin and inner organs, to enable the automatic computation of performance evaluation measures based on the simulation of all potential complications in accordance with the situations that may be encountered in the real clinical setting, as will be described more in detail later in this paper. The first step in achieving a visually realistic and physically accurate paracentesis simulation is to develop models for the deformable tissues and anatomical structures of the human body involved in this invasive procedure. Elastic deformation modeling of soft tissues is a wide research area in the field of 3D computer graphics and animation, with many applications particularly in VR-based systems. The two most widely employed generic methodologies are based on: (a) mass-spring models, where the deformable object is modeled as a lattice of point masses interconnected by spring/damping elements; and (b) finite element (FE) models, which are based on the discretization of a continuum elasticity model (Gibson & Mirtich, 1997). The use of FE models is typically preferred for surgical simulation applications (Bro-Nielsen, 1998; DiMaio et al., 2003), mainly because these models are tuned more intuitively than mass-spring nets. However, to enable real-time interaction with 3D FE models, long preprocessing steps are required, imposing additional constraints related to shape changes and deformation limits. The use of “intelligent model simplifications” for particular applications may overcome these problems in a task-specific context. In this framework, we have developed a simplified 2D (planar) finite element model to simulate deformation of the skin under point load. More details on this approach can be found in Tzafestas, Koumpouros, and Birbas (2004a, 2004b). It must be pointed out here, though, that in the pilot study presented in this paper, visual rendering of the skin deformation during needle insertion was disabled, in order to eliminate any force-substitution effects that this may add to the interactive simulation (since our focus in this study, as will be more clearly described in the rest of the paper, was on the effect of a direct haptic component). Another important issue here is real-time collision detection between moving objects in the virtual scene. In the first prototype platform used during the pilot study of this paper, the necessary routines have been implemented based on the “ColDet” 3D collision detection library for generic polyhedra, which is freely available on the Web. 1 The 3D surface models of the human body (outer skin), as well as those for the different anatomical structures (inner organs) involved (venal tree, muscle groups, lungs, and bony structures), are adapted from commercially available 3D models of human anatomy, 2 which are imported in an OpenGL/ GLuT application using a specially developed loader for obj (standard ascii) formatted files (for portability reasons). The most important issue in a haptic display system is undoubtedly that of achieving an optimal trade- off between two, often contradictory, requirements: transparency and stability. The first one is related to the realism of the simulation and the physically-based accuracy of the forces displayed to the user, which calls for complex and usually computationally expensive models. The second requirement calls for efficient calculations and real-time fast control loops to ensure a stable force- reflecting interaction. The main bottleneck in this respect is the visualization loop, performing all the necessary computations regarding collision detection, contact, and deformation modeling within the 3D graphics environment. These procedures all together usually run at a 20 –30 Hz frequency, which is definitely a very slow update rate with respect to the real-time control requirements. The force feedback control loop usually operates in a separate thread at 1 KHz or more. However, this fast control rate is in fact not exploited, if all critical updates for force-feedback computation (contact location, deformation data, etc.) are directly coming from a slow graphics loop. Similar to the effect of a ze- ro-order hold in any control system, this delay in ob- taining new critical information updates may cause undesirable chattering or even instability, particularly when simulation of hard contact is involved. All these problems are now very well known in the haptics research community, and several solutions have been proposed, most of them related to the application of some type of a virtual coupling (instead of a hard direct force/position interconnection) aiming to improve stability of interaction between haptic master and virtual simulation environment, for example, Colgate, Stanley, and Brown (1995) and Mahvash and Hayward (2005). It is needless to say, though, that such a soft coupling between any master and slave system may deteriorate the transparency of the system and, in our case, significantly decrease the realism of the simulation with respect to the haptic skills involved in the specific simulated procedure. In this respect, we have decided to employ, instead, a local geometry approximation technique, decoupling the haptic display computations from the visualization engine, for the particular case of a needle in contact with specific anatomical structures (such as the flat-type human skin or the cylindrical-type clavicle bone). Feel- ing these constraints through the sense of touch (particularly the hard haptic constraint invoked by the pres- ence of the clavicle bone above the needle track), by means of the forces exerted on the manipulated needle, is considered as the most important part of the manual skill involved in performing paracentesis of the subclavian vein. Ensuring efficient computations and fast updates was considered of primary importance, even if it is to the disadvantage of the accuracy of the geometrical representations used in force feedback. This was also subsequently validated during experimental trials through qualitative assessments performed by experienced surgeons. Such local force models, or lineariza- tion techniques, have also been proposed in other similar contexts, for example, in Kim, Otaduy, Lin, and Manocha (2003). Of course, different approximation models need to be applied for different simulation scenarios, also keeping in mind the particular requirements of the haptic skills involved in each specific application. Regarding now the needle reaction forces applied by the skin along the needle insertion direction, we assume inhomogeneity characteristics in depth (stiffness coefficients increasing with the deformation magnitude u ), resulting in a nonlinear stiffness coefficient that follows a typical exponential ...