Measuring in Virtual Reality: A Case Study in Dentistry

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 6, JUNE 2008 1177 Measuring in Virtual Reality: A Case Study in Dentistry Desmond M. Germans, Hans J. W. Spoelder, Luc Renambot, Henri E. Bal, Sander van Daatselaar, and Paul van der Stelt Abstract—For application specialists to accept virtual reality (VR) as a valid new measuring environment, we conducted several case studies. This paper describes a case study in dentistry in which VR is used to measure the length of the root canal of a tooth from processed tomography data. This paper shows the advantages of measuring in visualization space rather than in reality. We present a software framework aimed at the application specialist rather than the VR expert. Furthermore, the VR measuring technique is analyzed and compared to traditional measuring techniques for this particular application. We show that VR allows for intuitive measuring paradigms that are accurate and versatile alternatives to situations where traditional techniques are deficient. Index Terms—Computed tomography (CT), imaging, measuring, virtual reality (VR), visualization. I. I NTRODUCTION HE CURRENT trends in virtual reality (VR) show that interactive and collaborative applications are key research topics [19]. In particular, the integration of interactive VR with data visualization software is an important issue within the field [14], [18], [22]. Toward complete utilization of the potential of VR, one type of interaction is of particular importance to application specialists, i.e., quantitative measurement in visualization space. By using VR techniques to measure in the domain of the visualized data, more flexibility is offered for 3-D analysis [20]. The measurement environment and the data can be shared (simultaneously and in real time) among different scientists at various locations. Moreover, one can measure any type of derived quantity (not necessarily physically related) from the data at hand. We provide the means to interpret VR visualization of the data as a measuring environment by itself. T Manuscript received April 12, 2005; revised September 5, 2006. D. M. Germans is with the Division of Physics and Astronomy, Faculty of Sciences, Vrije Universiteit, 1081 Amsterdam, The Netherlands. H. J. W. Spoelder, deceased, was with the Division of Physics and Astronomy, Faculty of Sciences, Vrije Universiteit, 1081 Amsterdam, The Netherlands. L. Renambot was with the Division of Mathematics and Computer Science, Faculty of Sciences, Vrije Universiteit, 1081 Amsterdam, The Netherlands. He is now with the Electronic Visualization Laboratory, University of Illinois at Chicago, Chicago, IL 60607 USA. H. E. Bal is with the Division of Physics and Astronomy and the Division of Mathematics and Computer Science, Faculty of Sciences, Vrije Universiteit, 1081 Amsterdam, The Netherlands. S. van Daatselaar and P. van der Stelt are with the Department of Oral and Maxillofacial Radiology, Academic Center for Dentistry Amsterdam, 1066 Amsterdam, The Netherlands. Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TIM.2008.915952 This allows the scientist to intuitively look at the problem again, whereas the actual measuring (in the traditional sense) is more automatically performed in the data acquisition stages. To realize measuring in the visualization domain, a set of libraries [6] has been developed. These libraries allow application specialists to combine various aspects of VR and measuring into an application prototype using a simple application program interface (API). In addition, the data resulting from experiments or simulations can be measured, and correct approaches to the calibration and validity of measuring tools can be assessed. A real-world experiment in dentistry, where all the aforementioned aspects of measuring in VR are covered, is measuring the length of the root canal of a tooth. Traditionally, this is done by either physically measuring the length or analyzing an X-ray photograph of the tooth. Physical measurement is complicated because the tooth is embedded in the jaw. Furthermore, because the root canal is a 3-D structure, a projection to a conventional 2-D X-ray image can leave out essential information, which makes the analysis less reliable. Inspired by this problem, this paper shows the steps to take in setting up and calibrating VR as a measuring tool. In addition, we present a case study where measuring in VR is compared to traditional techniques and show that VR is a viable, valid, and flexible alternative to traditional measuring environments. The contributions of this paper are as follows. • We present a framework for measuring visualized data in VR, which is aimed at the application specialist. • We describe a case study in dentistry where this framework is used to measure the length of the root canal of a tooth. • We compare this technique with traditional measuring methods for this application. This paper is structured as follows. Section II shows related work, and Section III explains the idea of measuring in VR and explains the VR framework. Section IV presents the experimental setup to measure the length of the root canal. Section V explains the experiment and gives the results. Conclusions and future work are presented in Section VI. II. R ELATED W ORK In this paper, we use visualization in a VR environment, which is used to measure the length of the root canal of a tooth from computed tomography (CT) scans. Combining scientific visualization with VR could be done by adapting existing visualization packages like IBM’s OpenDX [11], AVS [4], or the Visualization Toolkit (VTK) [17]. The 0018-9456/$25.00 © 2008 IEEE 1178 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 6, JUNE 2008 visualization package can then be combined with scene graph libraries such as OpenGL Performer [15] or OpenSG [12] and VR frameworks like the CAVElibrary [3] or VR Juggler [7]. Scene graph libraries present the user with a hierarchical visualization tree, where each node can contain visualized data from the adapted visualization packages. VR frameworks handle tracker hardware, multidisplay environments, etc., for the final stages of the visualization. Many of these libraries have specific hardware requirements. For instance, OpenGL Performer is originally tailored for SGI machines. In addition, dealing with a set of modified libraries, each one covering part of the problem, results in a complex task for the application specialist (as opposed to a computer graphics specialist). The interaction with the virtual environment and the data is another feature required to conduct VR measuring experiments. A notable system providing visualization, scene graph support, and interaction is OpenInventor [23]. OpenInventor provides a scene graph hierarchy and ways to directly manipulate and edit objects in the visualization domain. However, originally, OpenInventor was targeted at desktop interaction instead of interactive VR. Examining the structure of the root canal of the tooth is done by recreating a 3-D model of the tooth from tomographic data [5], [13]. Here, high-resolution tomographic voxel geometries are examined to measure volumes, surfaces, and diameters. However, research on this does not include the use of interactive VR techniques to acquire measurement conditions and provide immediate feedback. In [1], Arnold et al. researched dealing with user interface issues when visualizing CT scan results. Their work combines surface representations of teeth with force feedback and collision detection to mimic the conditions under which dentists work. A considerable problem in the quantitative measurement of the root canal is the definition of the start and end of the canal. The addition of force feedback and collision detection might aid in this as the dentist gets a more natural feeling of what he is measuring. III. M EASURING One could say that measuring is a quantification of an observation in a given space. We consider a virtual environment that displays data as such a space. There are several advantages in using the visualization space for measuring. Primarily, quantities can be measured with great ease, even quantities that cannot be measured in reality or are commonly very costly to measure. Moreover, VR provides great flexibility in further prototyping the measurement experiments among multiple scientists at different sites. We will first look at the example of our case study. Fig. 1 shows an overview of measuring as applied to the case study. The top part shows the acquisition and reconstruction of the tooth data, and the bottom part shows the VR environment where the virtual measuring takes place. In this case study, a local CT is applied to obtain a series of radiographic images. These images are combined to reconstruct tomographic slices of the data. The slices make up a volume, and since the tooth is a solid object, an isosurface is extracted from this volume to Fig. 1. Process of measuring the root canal in VR. represent the tooth in VR. This tooth is visualized and presented to the user. The user can now directly measure the length of the root canal of the tooth from the visualization domain and can adjust parameters for slice reconstruction (filtering) and isosurface extraction (threshold). These parameters are then fed back into the process to generate a new representation of the tooth. A. Calibration To qualify as a valid measuring technique, the VR application must be calibrated. There are two ways to achieve this. The simplest way is to measure an object of known dimensions (a golden standard) and compare this to each measured object. The advantage of this method is that only one single calibration is required. However, all measurements should be done under identical conditions, so the filtering and threshold values may not change. A more elaborate way is to individually calibrate every step (charge-coupled device (CCD) camera, reconstruction, isosurface extraction) of the setup and work with a combined calibration over all steps. This can introduce errors but is more flexible, leaving the user to freely experiment with filtering and GERMANS et al.: MEASURING IN VIRTUAL REALITY: A CASE STUDY IN DENTISTRY threshold values. In this paper, we will use the first method and assume that the filtering and threshold values are fixed over a measurement series. B. VR Framework An application expert should rely on the simple and clear functionality of one or more compatible libraries. These libraries provide communication, measurement paradigms, and interactive VR primitives. With very little programming effort, the programmer then creates an interactive VR application that suits the needs of the field of research at hand. To develop such a set of libraries, three main levels can be identified. The first level deals with platform independence, basic graphical primitives, and interactive features. We developed a library (i.e., Aura) that provides these basic features. It presents a simple C++ API to issues like scene graph management and accessing input devices (trackers, keyboard, mouse, etc.). The second level is the presentation of measuring tools, manipulation primitives, and other ways to interact on a high level with the visualized data. On top of Aura, we developed a library [6] [i.e., Virtual Immersive Reality Program Interface (VIRPI)] to facilitate this. The third level is the communication between the various independent parts of the experiment, the virtual environment, and the data sets on which measurement is performed. For this, we developed CAVEStudy [10]. CAVEStudy wraps a remotely running program or repository and presents an API to the application specialist. Aura: As previously mentioned, Aura presents a lightweight C++ API to the underlying low-level graphical and interactive issues. The core of the graphical part of Aura consists, like OpenGL Performer, of nodes in a scene graph. These nodes can be geometry nodes, cameras, lights, etc. Next to this, Aura defines a variety of simple shapes (cube, ball, cylinder, arrow, pyramid, to name a few) and can load graph data (3-D model files from modelers and other polygon descriptions), fonts, and textures/images. Aura can encapsulate CAVElibrary or VR Juggler, providing a seamless interface within the scene graph hierarchy to input devices and rendering contexts. On traditional workstations, a simulator (much like the CAVElibrary simulator) fills in the missing hardware. In the current implementation, Aura comes as a set of libraries for various setups on different platforms (IRIX, Linux, and Windows). Aura is functionally the same for each setup on each platform. When the given setup is selected, simply creating the Aura environment object initializes the necessary hardware and tools. To reduce complexity, Aura does not present issues like multipipe output, multiprocessor systems, and shared memory to the programmer. VIRPI: The high-level VIRPI toolkit rests on top of Aura and is identical for all platforms. VIRPI is roughly based on concepts and ideas from 2-D GUIs like X, Qt, or GTK, where events are passed across a tree structure. These events are typically external events, such as clicks of buttons on the pointer, movement of the pointer, or other trackers, keystrokes, and joystick changes. 1179 Next to the basic event tree structure, VIRPI provides several simple controls to interact with the user. The functional parts of 2-D GUIs, like menus, sliders, and (radio) buttons, have a VR counterpart in VIRPI but are designed to operate in VR with a limited set of inputs. Visual examination and selection are important aspects beneficial to the VR user. These tasks are generally done with manipulators, i.e., constructs that interpret events from the user to make an object move, rotate, or scale in an intuitive way. Several standard manipulators are provided in a similar fashion as with OpenInventor. To define a measurement, the programmer is given a series of simple classes to adjust settings, display values, and move selections in the data set. Depending on the measurement at hand, the user can define volumes, planes, or points in the data set. For our case study, the user defines the control points of a flexible spline yardstick by moving them around in VR. CAVEStudy: To minimize programming of the control over the data set generation programs, the user has to describe the program with a description file. This file is processed by CAVEStudy to generate two objects, i.e., a proxy and a server object (see Fig. 1). The program is wrapped into a server object to control its execution. This way, the program stays in its native execution environment, and no alterations to it are needed. The server’s interface provides methods to start, stop, pause, and resume the program. The data generated by the program is automatically propagated to the proxy object using the CAVERNSoft [9] network layer. The proxy object can be seen as a local copy of the remote program. Through the network, it relays the input values and the commands to the server. Furthermore, it manages the incoming data from the simulation and presents it to VIRPI. By using CAVERNsoft, it is possible to access one simulation with multiple VR setups. This way, a basic collaboration setup can be realized among multiple sites. Each site can, depending on their VR setup and the individual wishes of the users, display different representations of the data. IV. C ASE S TUDY Using Aura, VIRPI, CAVEStudy, and proprietary programs to generate the tooth representation, Fig. 1 is realized. To calibrate the setup and assess the usefulness of VR and the correctness of the results, an isolated tooth is measured in three ways: 1) physically measuring the endodontic file with a caliper; 2) traditionally measuring on a projected X-ray picture (the common dentistry method); 3) VR using the new VR method. A. CT Scanner To get an accurate scan of one of the patient’s teeth (or any other region in the maxillofacial area), an imaging procedure based on CT (called Local CT) is used. This essentially means that a series of X-ray projections are made at different angles 1180 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 6, JUNE 2008 Fig. 2. Setup used to measure an individual tooth. Note that to get exposures at different angles, the tooth is rotated in the setup. around the patient. These “images” are captured with a CCD device and processed to reconstruct the volume. In a conventional CT, the X-ray beam used is wide enough to cover the whole width of the patient. In Local CT, the beam is much smaller and only covers the region of interest (ROI), which results in much more favorable dose conditions. For the case study described here, the Local CT is used, and the ROI is a molar, which is much smaller than the head of the patient. Using a narrow beam means that the volume is only partially sampled, and the projections of the ROI are corrupted by structures surrounding the ROI, e.g., the other teeth in the mouth, and the hard structures of the maxillofacial region. To reconstruct the volume, the standard filtered backprojection CT algorithm [8] is still usable although the surrounding tissue will reduce the contrast. The experimental setup consists of an X-ray source, an object table, and a detector mounted on an optical bench (Fig. 2). The X-ray source is a standard device (commonly used in dental practice), and it produces a conal beam. However, the current setup has a focus-to-object distance that is large with respect to the object-to-sensor distance. Therefore, we approximate the conal beam with a parallel beam. We currently use a Sirona Sidexis CCD detector (664 × 872 pixels with a 12-bit precision). B. Software The reconstruction of the slices is done by using an implementation of the filtered backprojection algorithm [21]. The projection images are filtered with a ramp filter and backprojected onto the slices. These slices are passed as volume data through an isosurface extraction program. This was programmed using the VTK library, which provides functionality to apply visualization primitives to a data set (the “Marching Cubes” algorithm, mesh decimation, etc.). Finally, the resulting triangle mesh is decimated. With Aura, VIRPI, and CAVEStudy, the reconstruction, isosurface generation, and visualization are merged. CAVEStudy retrieves the decimated triangle mesh for the VR application. Using the VIRPI framework, an application is built that displays a translucent tooth in a box (Fig. 3) and allows the user to Fig. 3. Overview of the VR measuring environment. The tooth reconstruction is suspended in a box (indicated by cylinders here), which can be used to rotate the tooth. The spline yardstick is shown in the tooth. To the left, the current length of the yardstick is displayed. rotate the box around its three main axes. This way, the user can investigate the tooth data from various angles. The box also contains a spline yardstick that is used to interactively measure the length of the root canal. The yardstick consists of four movable control points and a connected Catmull–Rom spline [2] between them. The Catmull–Rom spline is used for its continuous-tangent property. The tangents of the spline pieces at each control point are intrinsically defined in such a way that the spline is guaranteed to be a continuous curve. Because we wish to measure a curved trajectory, at least three control points are required. In our setup, four control points are used, which allows the user to specify a nonplanar curve. The actual measuring is done by calculating the length of this spline. To support the user in finding the root canal in the tooth, two features are added to the program. First, the tooth can be made transparent. This way, the user can get an overview of the interior of the tooth, including the root canals. Second, a clipping plane is used to make part of the tooth invisible when the user approaches it. This too gives the user a better overview of the tooth’s interior. The threshold value of the isosurface extraction and the filter settings for slice reconstruction are controlled by sliders in the virtual environment. From these sliders, variables are sent back to the reconstruction program. The current length, threshold value, and filter settings are displayed on a display board in the virtual environment. Fig. 4 shows the view of the virtual environment as the tooth is being measured. Being able to change the threshold value and the filter settings has a direct influence on the quality of the measurement. If the threshold value is taken too low, the visualized root canal is too long. If the threshold value is taken too high, the root canal is visualized too short or not visible at all. Similarly, setting incorrect filter parameters results in a poor reconstruction. C. Calibration The measurement procedure for the VR setup essentially consists of comparing the length of the spline placed by the user with the length of a “golden standard” or reference stick GERMANS et al.: MEASURING IN VIRTUAL REALITY: A CASE STUDY IN DENTISTRY Fig. 4. Side view of the measuring environment. Here, the second control point is being placed in the root canal. The length is continuously calculated so the indicator (here at the back) changes when a control point is moved. Fig. 5. Traditional measurement of the root canal length of each of the root canals in a separate tooth. The file that is photographed with the tooth is measured from the resulting X-ray image with a ruler. that is separately measured beforehand. In general, displaying 3-D graphics already implies a metric with a unit length. Thus, calibrating the VR setup as a measurement device comes down to finding a factor with which to scale the unit length already at hand. To do this as objectively as possible, an automatic calibration routine is designed. This routine takes the total set of slice images from the reconstruction program. Assuming that the stick is vertically placed on the optical bench, the routine finds the lowest and highest slice on which the stick is visible. From both slices, the center of gravity in the feature is calculated, thereby locating the center of the stick’s start and endpoints in the volume. Knowing these and knowing how far the slices are apart in the visualization, the length of the stick can be calculated. Dividing the length of the stick that was measured with a caliper (20.5 mm) by the length obtained from the above algorithm gives the calibration factor. The inaccuracies in this algorithm are due to noise in the CCD and the propagation effects of the filtered backprojection algorithm. We neglect these effects with respect to the error margins of the other measuring methods (physical and traditional) as they are much smaller (Fig. 5). 1181 Fig. 6. Distribution of lengths of the golden standard stick measured by several users. The first vertical line indicates the actual length of the golden standard stick. The second vertical line shows the average length as measured by the test users. V. E XPERIMENTS To assess the validity of using VR as a measuring environment, we let a group of test users measure two objects, i.e., the reference stick and one root canal of the tooth. Measuring the reference stick shows how the measurement error is related to the user interface and the ability of the user to correctly interpret what he sees. Measuring the tooth then gives the actual results and shows tendencies due to the user’s interpretations. Before the experiments are conducted, the reference stick is measured to be 20.5 ± 0.1 mm. Then, the filter settings and isosurface threshold are fixed, so the reconstruction of the stick clearly coincides with the slices. Finally, the automatic calibration algorithm is applied to yield the calibration factor for 3-D space. For the actual experiments, seven users are asked to measure the stick five times and the root canal three times. The measurements are done in the CAVE environment at the Academic Computing Services Amsterdam (SARA) [16] with an Ascension Flock-of-Birds tracking system. A. Reference Stick In VR, the interpretation of the location of the start and end of the root canal becomes important. The stick, as opposed to the tooth, is a clearly defined shape for which the start and end points provide no interpretational differences among the users. Because the stick is a cylinder with nonzero radius, the length of the cylinder is the smallest possible distance between the start and endcaps of the cylinder. All other connections between the start and endcaps yield a larger length. This implies that the measurements will likely show an overestimation rather than an underestimation. Fig. 6 shows a frequency distribution of the measured lengths. The average measured length is 20.57 mm with a standard deviation of 0.07 mm. As expected, there is an overestimation in the measurements. A second source for overestimation is shown in Figs. 7 and 8. This shows the actual Z-coordinates of the top and bottom control points of the spline yardstick for every user with respect to the start and endcaps of the stick. There is a tendency for users to avoid placing the control points inside the stick, thereby introducing the extra (small) overestimation. 1182 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 6, JUNE 2008 Fig. 7. Z-coordinates of the top control points on the golden standard stick. Each line indicates a user, whereas each glyph indicates a measurement. The horizontal line is the top as calculated by the calibration algorithm. The grey area around this line indicates a region corresponding to 1 cm of the actual deviation for the user’s tracker. Fig. 10. X- and Y -coordinates of the top control point on the tooth root canal. TABLE I ROOT CANAL MEASURED WITH THE THREE DESCRIBED METHODS Fig. 8. Z-coordinates of bottom control points on the golden standard stick. Each line again indicates a user, whereas each glyph indicates a measurement. The horizontal line is the bottom as calculated by the calibration algorithm. The grey area around this line indicates a region corresponding to 1 cm of the actual deviation from the user’s tracker. the definition of the center can introduce an overestimation. However, this does not need to be large, because the root canal is bent, and all control points influence the measured length. If the first segment in the spline can be approximated with a straight line, the trigonometry in Fig. 10 shows that the largest overestimation is on the order of 0.1 mm, which is one order of magnitude smaller than the error margins of the other methods. C. Results Fig. 9. Z-coordinates of the bottom control point on the tooth root canal. Each line indicates a user, whereas each glyph indicates a measurement. The horizontal line is the bottom as calculated similar to the calibration algorithm. The grey area around this line indicates a region corresponding to 1 cm of the actual deviation from the user’s tracker. Note that the tendency not to touch the tooth is higher than for the golden standard stick. B. Tooth Because the tooth is a less well-defined shape, it is interesting to know the interpretation issues of the start and end of the root canal. Fig. 9 shows the Z-coordinates of the last control point of the spline yardstick for each measurement for each user. The end of the root canal is rather well defined, and the measurements there show a similar effect as for the stick; users tend to avoid placing the bottom control point inside the tooth. Fig. 10 shows the spread of the top control points on a plane perpendicular to the vertical axis through the center of the tooth. Like for the cylindrical shape of the reference stick, The measurements of the test users on the root canal are averaged to come to the result in Table I. We see that the traditional method systematically indicates shorter lengths than the other two methods. This shows that a projection of the tooth (an X-ray photograph) is not sufficient for accurate measuring. A projected length is always shorter than or at most equal to the real length. Furthermore, we see the suggested overestimation in the VR method due to the interpretational and user interface issues described above. VI. C ONCLUSION This paper has shown that it is possible to use VR as an experimentation environment where real-world measuring paradigms can be applied. Both software technical and measurement issues that are addressed present no insurmountable drawbacks. It is possible to create, validate, and calibrate a measurement GERMANS et al.: MEASURING IN VIRTUAL REALITY: A CASE STUDY IN DENTISTRY experiment using the Aura/VIRPI and CAVEStudy libraries and proprietary software-translating information from the physical experimental setup. Apart from measuring processed real-world data, measuring in VR opens possibilities to measure highly complex data sets, where classical measuring is not possible, too expensive, or hard to accurately realize. Visualizing the derived information presents new quantities to observe and measure. Furthermore, this paper has shown a novel noninvasive way of measuring the length of the root canal of a tooth of a patient subject to a minimal dose of radiation. The method yields accurate results and shows that the traditional use of a 2-D X-ray projection is not sufficient. As noted before, because the user has a very flexible control over the measuring process, the user can analyze very noisy input data, which implies that the radiation dose could be reduced even further. Toward improving the technique and exploring further what measuring in VR has to offer, we will extend work on the collaborative aspects of the system. Scientists can greatly benefit from the availability of a collaborative experimentation and analysis environment in which they can discuss findings with peers around the world. Next to this, we plan to do more usability tests and further improve the calibration process for the case study at hand. The local CT method used in the case study is currently applied to a separate tooth on a conditioned optical bench. Further research in the method will be done using a tooth in a jaw, a tooth in a jaw of a dosimetric dummy (phantom head), and, finally, a real patient. R EFERENCES [1] D. B. Arnold, A. M. Day, M. R. Goetz, A. Courtenay, and H. I. Graham, “Virtual teeth for endodontics training and practice,” in Proc. Int. Conf. IV, 2000, pp. 597–604. [2] E. Catmull and R. Rom, “A class of local interpolating splines,” in Computer Aided Geometric Design. New York: Academic, 1974, pp. 317–326. [3] C. Cruz-Neira, D. J. Sandin, and T. A. DeFanti, “Surround-screen projection-based virtual reality: The design and implementation of the CAVE,” in Proc. SIGGRAPH, J. T. Kajiya, Ed., Aug. 1993, vol. 27, pp. 135–142. [4] I. Curington and M. Coutant, “AVS: A flexible interactive distributed environment for scientific visualization applications,” in Proc. 2nd Eurographics Workshop Scientific Visualization, 1991. [5] S. E. P. Dowker, G. R. Davis, and J. C. Elliot, “Nondestructive threedimensional imaging for in Vitro endodontic studies,” Oral Surg. Oral Med. Oral Pathol., vol. 83, no. 4, pp. 510–516, Apr. 1997. [6] D. M. Germans, H. J. W. Spoelder, L. Renambot, and H. E. Bal, “VIRPI: A high-level toolkit for interactive scientific visualization in virtual reality,” in Proc. Immersive Projection Technol. Virtual Environments, May 2001, pp. 109–120. [7] C. Just, A. Bierbaum, A. Baker, and C. Cruz-Neira, “VR Juggler: A framework for virtual reality development,” in Proc. 2nd IPT Workshop, 1998, pp. 1–8. [8] A. C. Kak and M. Slaney, Principles of Computerized Tomographic Imaging. Piscataway, NJ: IEEE Press, 1988. [9] J. Leigh, A. E. Johnson, T. A. DeFanti, and M. Brown, “A review of teleimmersive applications in the CAVE research network,” in Proc. IEEE Virtual Reality, 1999, pp. 180–187. [10] L. Renambot, H. E. Bal, D. M. Germans, and H. J. W. Spoelder, “Cavestudy: An infrastructure for computational steering in virtual reality environments,” in Proc. 9th IEEE Int. Symp. High Performance Distrib. Comput., Pittsburgh, PA, Aug. 2000, pp. 57–61. [11] B. Lucas, G. Abram, N. Collins, D. Epstein, D. Gresh, and K. McAuliffe, “An architecture for a scientific visualization system,” in Proc. IEEE Vis. Conf., 1992, pp. 107–114. 1183 [12] OpenSG, Portable opensource scene graph system. [Online]. Available: http://www.opensg.org [13] O. A. Peters, A. Laib, P. Ruegsegger, and F. Barbakow, “Threedimensional analysis of root canal geometry by high-resolution computed tomography,” J. Dent. Res., vol. 79, no. 6, pp. 1405–1409, Jun. 2000. [14] B. Plale, G. Eisenhauer, K. Schwan, J. Heiner, V. Martin, and J. Vetter, “From interactive applications to distributed laboratories,” IEEE Concurrency, vol. 6, no. 2, pp. 78–90, Apr.–Jun. 1998. [15] J. Rohlf and J. Helman, “IRIS performer: A high performance multiprocessing toolkit for real-time 3D graphics,” in Proc. SIGGRAPH, Annu. Conf. Series, A. Glassner, Ed., Jul. 1994, pp. 381–395. [16] SARA, Academic Computing Services Amsterdam. [Online]. Available: http://www.sara.nl [17] W. J. Schroeder, K. M. Martin, and W. E. Lorensen, “The design and implementation of an object-oriented toolkit for 3D graphics and visualization,” in Proc. Visualization, R. Yagel and G. M. Nielson, Eds., San Francisco, CA, Oct. 27–Nov. 1, 1996, pp. 516–517. 93-100, 472. [18] C. Shaw, M. Green, J. Liang, and Y. Sun, “Decoupled simulation in virtual reality with the MR Toolkit,” ACM Trans. Inf. Syst., vol. 11, no. 3, pp. 287–317, Jul. 1993. [19] S. Singhal and M. Zyda, Networked Virtual Environments: Design and Implementation. Reading, MA: Addison-Wesley, 1999. [20] H. J. W. Spoelder, “Virtual instrumentation and virtual environments,” IEEE Instrum. Meas. Mag., vol. 2, no. 3, pp. 14–19, Sep. 1998. [21] A. N. van Daatselaar, S. M. Dunn, H. J. W. Spoelder, D. M. Germans, L. Renambot, H. E. Bal, and P. F. van der Stelt, “Feasibility of local CT of dental tissues,” Dentomaxillofacial Radiol., vol. 32, pp. 173–180, 2003. [22] A. van Dam, A. S. Forsberg, D. H. Laidlaw, J. J. LaViola, Jr., and R. M. Simpson, “Immersive VR for scientific visualization: A progress report,” IEEE Comput. Graph. Appl., vol. 20, no. 6, pp. 26–52, Nov./Dec. 2000. [23] J. Wernecke and I. Mentor, Programming Object-Oriented 3D Graphics With OpenInventor. Reading, MA: Addison-Wesley, 1994. Desmond M. Germans received the M.Sc. degree in physics from Vrije Universiteit, Amsterdam, The Netherlands, in 1998. He is currently working toward the Ph.D. degree at the Physics Applied Computer Science Group, Division of Physics and Astronomy, Vrije Universiteit. He has worked on several projects concerning virtual reality and applying interactive visualization to medical and physics research. His main interests cover visualization and 3-D graphics, virtual reality, and interactive systems. In 2003, he started a consultancy company for media technology. Hans J. W. Spoelder received the M.Sc. degree in physics and the Ph.D. degree in biophysics from Vrije Universiteit, Amsterdam, The Netherlands, in 1980 and 1987, respectively. He has researched many topics regarding the boundaries of physics and computer science both locally and with the Verdical User Environment Group, IBM T. J. Watson Research Center, Hawthorne, NY, which earlier included computational physics and later visualization and the integration of education and information systems with the Physics Applied Computer Science Group that he started. His last projects included the ICWall, which is a stereo-tiled display placed in a classroom to enable education supported by stereo 3-D graphics. He passed away on April 1, 2003. Luc Renambot received the Ph.D. degree in computer science from the INRIA Research Institute, Universite de Rennes 1, Rennes, France, in 2000. In 2000, he developed the CAVEStudy system in Amsterdam, The Netherlands. His work includes interconnecting virtual reality systems and grid computing. He is currently a Postdoctoral Researcher with the Electronic Visualization Laboratory, University of Illinois at Chicago. 1184 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 6, JUNE 2008 Henri E. Bal received the M.Sc. degree in mathematics from Delft University of Technology, Delft, The Netherlands, in 1982 and the Ph.D. degree in computer science from Vrije Universiteit Amsterdam (VUAmsterdam), Amsterdam, The Netherlands, in 1989. He is currently a Full Professor with the Faculty of Sciences, VUAmsterdam, where he heads a research group on parallel programming. He is the author of Programming Distributed Systems (Prentice-Hall, 1991) and the coauthor of Programming Language Essentials (Addison-Wesley, 1994) and Modern Compiler Design (Wiley, 2000). His research interests include parallel and distributed programming and applications, grid computing, interactive applications, and programming languages. Dr. Bal was the Program Chair of the 2nd IEEE International Symposium on Cluster Computing and the Grid (CCGrid 2002) and the Program Cochair of the 15th IEEE International Symposium on High-Performance Distributed Computing (HPDC-15). He is the Adjunct Director of the Dutch “Virtual Laboratories for e-Science” (VL-e) project. Sander van Daatselaar received the M.Sc. degree in physics from Vrije Universiteit, Amsterdam, The Netherlands, and the Ph.D. degree from the Academic Center for Dentistry Amsterdam (ACTA), Amsterdam. He is currently with the Department of Oral and Maxillofacial Radiology, ACTA. His interests include medical systems and the Local CT setup at ACTA. Paul van der Stelt received his training as Dentist and Oral Radiologist. He is currently the Chairman of the Department of Oral and Maxillofacial Radiology, Academic Center for Dentistry Amsterdam, Amsterdam, The Netherlands. His main research interest is digital radiology.