Comparison of Tracking Techniques for Intraoperative Presentation of Medical Data using a See-Through Head-Mounted Display Tobias Salb, Oliver Burgert, Tilo Gockel, Björn Giesler and Rüdiger Dillmann Industrial Applications of Informatics and Microsystms (IAIM) Building 07.21, Department for Computer Science Universität Karlsruhe (TH), 76128 Karlsruhe, Germany Tel.: ++49 721 608 7126, Fax: ++49 721 608 8270 Email:
[email protected], WWW: http://wwwiaim.ira.uka.de/users/salb/
Tracking of a see-through head-mounted display is a necessary precondition for proper overlay of virtual data and real scenes within the display. In our contribution ,the intention and technique for Intraoperative Presentation will be presented. Focus will be the tracking of the display device. We will illustrate and compare three different optical tracking approaches and the results achieved by using them.
1 Introduction and purpose of the work Nowadays many tools for preoperative planning and simulation of surgical interventions are available in the clinical environment, while the surgical procedure itself still lacks the computer based assistance. At the Institute for Industrial Applications of Informatics and Microsystems (IAIM) at Universität Karlsruhe (TH), Germany, we are working on a solution for closing this gap using augmented reality. A technique called Intraoperative Presentation has been developed in order to superimpose virtual computer-generated information with a real patient using a see-through head-mounted display [2]. The purpose of our development is to visualise results from preoperative work, like planning and simulation information, radiological data, risk regions, target areas or any other important data within the see-through glasses. Clinical tests have been done with the display to ensure proper functionality and to receive considerations for the design of the visualisation software [1]. The concept of Intraoperative Presentation has been compared to other solutions [3] and was described in detail on the SPIE 2000 conference [2]. In this paper our focus is on the tracking of the see-through head-mounted display. Precise real-time tracking is necessary for achieving sufficient results in data presentation. Three different optical tracking algorithms have been implemented at our institute. Methods and results are being presented below.
Figure 1: Start GUI for Intraoperative Presentation, data for clinical trials of the display, virtual information that shall be visualised within the glasses (from the left to the right)
2 Methods Intraoperative Presentation of image data causes hard accuracy requirements. It is challenging to track the see-through glasses with sufficient precision in order to transform the image data correctly and to supply the surgeon with an optimal superimposition of virtual and real data. The first tracking method we call a “standard machine vision approach” using two Pulnix monochrome CCD cameras. Light emitting diodes are mounted on top of the display as artificial landmarks. The cameras are fixed in the operation room for tracking the display. For the tracking software, the Matrox Imaging Library has been used as a basis. Besides this standard setup, two miniature cameras will be mounted on both sides of the display in order to enhance precision of the tracking. The second tracking system is the commercial optical navigation system “Polaris”. The system, manufactured by NDI, is distributed by the Surgical Navigation Network (SNN). Results of the SNN tracking software have been combined with our visualisation system. Moreover, special rigid bodies with passive markers have been designed for proper tracking results. The last tracking method we are realising is an innovative tracking approach using a “parabolic mirror” which has been used in mobile robotics before. Artificial landmarks are distributed all over the operation room. Images are being taken using a single CCD camera attached on top of the head-mounted display. A parabolic mirror which is coated on the outside is mounted on the camera. This setup enables us to take a 360 degree image of an operation room. Using this type of images we search for the landmarks in order to do dynamic localisation and navigation of the display. The approach is still under development and will be presented in a separate paper in detail.
Figure 2: The different optical tracking approaches: Standard camera-based setup, NDI Polaris system and parabolic mirror approach (from the left to the right)
3 Results All tracking setups show sufficient time behaviour. While the Polaris system has been developed for real-time usage, the two self-made approaches had to be tuned properly. Fast PCs are being used in order to ensure real-time tracking. Because of the single camera within the setup, the parabolic mirror approach shows a better time performance. First clinical trials have been done for testing accuracy. After registration, the particular tracking bodies of the different approaches have been placed in measured positions and orientations. Then the results of the respective tracking software have been analysed with regard to the measured data. The SNN Polaris tracking has been able to reach a sub-millimetre accuracy in translation and a sub-degree accuracy in rotation. This high accuracy has been proposed by the company before. Our self-made camera-based approach has been worse in the test, with an accuracy of about 2-3 millimetres and about 2 degrees. The parabolic mirror approach has shown a good rotational behaviour in first trials while translation behaviour is worst. No Setup type Translation accuracy 1 Standard camera-based =< 3mm 2 SNN Polaris < 0,3 mm 3 Parabolic mirror =< 5 mm
Rotation accuracy =< 2,5 degrees < 1 degree =< 1,5 degrees
Figure 3: Accuracy test results
4 Conclusion For use in Intraoperative Presentation we have developed and compared three optical tracking systems. Best behaviour has been shown by an enhanced setup, based on the commercially available Polaris system. Much worse in translation but comparable in rotational behaviour are our self-made methods, using a parabolic mirror or a standard camera-based approach. Results are nevertheless satisfying and visualisation of data will be realised soon using the different tracking systems.
References [1] Brief, J., Hassfeld, S., Salb, T., Burgert, O., Münchenberg, J., Pernozzoli, A., Grabowski, H., Redlich, T., Raczkowsky, J., Krempien, R., Kotrikova, B., Wörn, H., Dillmann, R., Mühling, J. and Ziegler, C.: Clinical evaluation of a see-through display for Intraoperative Presentation of planning data. Proceedings of conference: Israeli Symposium on Computer-Integrated Surgery, Medical Robotics and Medical Imaging (ISRACAS), Haifa, May 2000. [2] Salb, T., Brief, J., Burgert, O., Hassfeld, S. and Dillmann, R.: Intraoperative presentation of surgical planning and simulation results using a stereoscopic see-through head-mounted display. Proceedings of conference: Stereoscopic Displays and Applications, Part of Photonics West (SPIE), San Jose, CA, January 2000. [3] Salb, T., Brief, J., Burgert, O., Hassfeld, S., Mühling, J. and Dillmann, R.: Towards augmented reality in medicine - Overview and approach. Proceedings of workshop: Computer Aided Surgery (CAS), Erlangen, October 1999.
