Development and validation of a three dimensional ultrasound based navigation system for tumor resection

Transcription

1 Repository of the Max Delbrück Center for Molecular Medicine (MDC) Berlin (Germany) Development and validation of a three dimensional ultrasound based navigation system for tumor resection S.S. Chopra, M. Hünerbein, S. Eulenstein, T. Lange, P.M. Schlag, and S. Beller Published in final edited form as: European Journal of Surgical Oncology (EJSO) Apr ; 34(4): doi: /j.ejso Elsevier (The Netherlands)

2 Final Draft Development and validation of a three dimensional ultrasound based navigation system for tumor resection S. S. Chopra 1, M. Hünerbein 1, S. Eulenstein 1, T. Lange 1, P. M. Schlag 1, and S. Beller 1 1 Department of Surgery and Surgical Oncology, Charité Campus Buch, Universitätsmedizin Berlin, Berlin, Germany ABSTRACT Background: Intraoperative navigation is a rapidly emerging procedure in orthopaedic surgery and neurosurgery. For abdominal tumors (e.g. liver metastasis) and soft tissue tumors there is only limited experience with navigation techniques due to problems of organ shift and tissue deformation. We have developed a navigation system for tumor resection in soft tissue based on 3D ultrasound imaging and optical tracking. Methods: Two different modes of navigation were evaluated and compared with conventional surgery in an experimental soft tissue model. Both techniques were based on 3D ultrasound and an optical tracking system for intraoperative real time registration of surgical instruments. These two techniques were used: a) Indirect navigation with ultrasound guided insertion of a tracked hook needle into the tumor; and b) Direct navigation using a 3D image which was obtained with an optically tracked 3D ultrasound probe. It was the aim of both techniques to achieve a circumferential resection margin of 2 cm around the tumor. Results: A total of 23 resections were performed consisting of indirect (n = 7) and direct (n = 10) navigation and conventional surgery (n = 6) as gold standard. For indirect navigation a median deviation from the ideal resection margin (accuracy) of 0.32 cm was measured. Direct navigation showed an accuracy of 0.16 cm compared to 0.42 cm with conventional surgery. Navigated surgery showed for both techniques a significant increase of resection accuracy compared to conventional resection (p < 0.05). Conclusion: 3D ultrasound based indirect and direct optoelectronic navigation for resection of soft tissue tumors is feasible and may improve intraoperative orientation with increased surgical precision. KEYWORDS Navigation; 3D Ultrasound; Optical Tracking; Soft Tissue Tumors. Introduction Image guided surgery with navigation techniques is an evolving procedure in neurosurgery, otolaryngology and orthopaedic surgery. [1,2] These systems improve intraoperative orientation and facilitate increased accuracy of tumor localization or bone resection. [2] Usually a preoperative three dimensional (3D) image data set (CT or MRI) is acquired and transferred to a workstation for post processing (planning).[3] During surgery the obtained image volume is referenced with the actual patient position (registration) and tracked surgical instruments are virtually displayed in their current spatial position. A problem of intraoperative manipulation is that the current anatomical situation is increasingly changed (e.g. organ shift) reducing the value of the preoperative image data. Publications on neurosurgery report inaccuracies due to brain shift of more than one centimetre. [4] It is therefore necessary to obtain updated intraoperative image volumes with intraoperative CT or MRI scans. [4,5] The implementation of these machines in the operation room is complicated and costly and data acquisition is time consuming. The operative therapy of abdominal tumors and soft tissue tumors is even more challenging because of an increased tissue shift, breathing artefacts and absent or reduced anatomical landmarks. [6] We have developed a navigation system for image guided surgery and evaluated two different 3D ultrasound based navigation techniques for the resection of soft tissue tumors. Both techniques were compared in an experimental setting to conventional surgery. Materials and methods Navigation system The navigation system consisted of a 3D ultrasound machine with an attached 3D abdominal ultrasound probe (Voluson 730 ; Voluson Abdominal transducer RAB 4 8 (P) GE Healthcare, Milwaukee, WI, USA) and an infraredbased optical tracking system (Polaris by Northern Digital Inc., Canada). For our experimental study we created a soft tissue tumor model by embedding a 1 cm [3] silicon phantom into the centre of a tissue specimen. Surgery was performed with an electrocauter (ValleyLab Inc., Boulder, Colorado, USA) under image guidance. Conventional surgery based on palpation was used as gold standard. The ideal resection margin was specified with 2 cm distance from the tumor surface as an arbitrary value. Indirect navigation Initially the tumor was localized by using a 3D ultrasound scan. Under ultrasound guidance a thermoablation needle (RITA Medical Systems Inc., Fremont, CA, USA) was introduced and fixed with the tip at the centre of the tumor. A passive tracking device equipped with retroreflective spheres (tracker) was attached at the distant end of the needle. The Polaris optical tracking system was used for continuous measurement of needle and instrument position. Both tools and the virtual tumor were displayed on a split screen in two orthogonal planes. The coronal plane visualized the electrocauter in relation to the resection margin while the longitudinal plane showed the depth of the instrument inside the tissue (Fig.1). In order to calibrate the guidance needle and the electrosurgical instrument we implemented the algorithm described in refs. [7,8]. During surgery the tumor and the instruments MDC Repository 1

3 were continuously virtually displayed in real time. A total of 7 indirect navigations were performed. Direct navigation For the localization and intraoperative registration of the tumor a tracker was attached to the ultrasound probe. The position of the probe was tracked by the Polaris system during acquisition of the 3D ultrasound data. Due to the preoperative calibration of the tracker position relative to the ultrasound image coordinates, these image coordinates were registered to the physical space of the phantom model. Retroreflective spheres were also attached to the electrocauter, and were calibrated and registered. The position of the tracked instrument was localized on the monitor and the instrument was navigated in the virtual environment of the registered 3D ultrasound data (Fig.2). Virtual orientation was achieved on two section planes. The coronal plane was in line with the ultrasound probe and the axial plane was oriented perpendicular to this, mostly in the transverse direction. A total of 10 ultrasound based navigations were performed. Evaluation The resected specimen were analyzed by helical CT imaging (GE Healthcare, Milwaukee, WI, USA) with a collimation of 1 mm, 1.5 pitch and a reconstruction increment of 2 mm was used. The images were transferred from the local network to a separate workstation and surface and 3D models were created with visualization software (Amira, Mercury Computer Systems Inc., USA). The most central plane of the tumor was selected and the resection margin marked with red points every 2 mm along the tumor surface (Fig.3). In the next step the distance to the tumor surface was measured for each point. The accuracy was defined as the unsigned distance to the security margin (deviations outside and inside of the security margin were considered equal). Statistical analysis The data were analyzed with the SPSS for Windows Software (SPSS Inc., Chicago, IL, USA) and displayed as boxplots. The absolute minimum and maximum values were shown as whiskers; the upper and lower box borders showed the 75% and 25% quartile and the black bar represented the median in contrast to the accuracy. The differences between the boxplots were established with the Mann Whitney U-test, whereby p values <0,05 were considered statistically significant. Results Two different modes of navigation were evaluated and compared to conventional surgery. An ideal resection margin of 2 cm was proposed. For indirect navigation a guiding needle had to be introduced at the tumor centre. This procedure took approximately 2 min and was always feasible. Indirect needle based navigation achieved a median accuracy of 0.32 cm (standard deviation 0.21 cm) with an average maximal and minimal margin between 2.54 cm and 1.7 cm. The duration of indirect navigated surgery was between 12 and 24 min with a median time of 18 min. During direct navigation, preoperative and intraoperative ultrasound scans of the tumor were obtained. The time required for acquisition of the scans and data postprocessing was 2 min. All tumors were correctly identified and navigated resection was always successful. Direct 3D ultrasound guided navigation produced a median accuracy of 0.16 cm (standard deviation 0.13 cm) with an average maximal and minimal margin between 2.51 cm and 1.81 cm. As a gold standard conventional surgery based on palpation showed a median accuracy of 0.42 cm (standard deviation 0.29 cm) with maximal and minimal margins between 2.79 cm and 1.06 cm. The detailed data are shown below (Table 1). The median accuracy of direct and indirect navigation of 0.16 cm and 0.32 cm was significantly higher (p < 0,05) than that of conventional resection (0.42 cm) (Fig. 4). However, it must be emphasized that the tumor was poorly palpable. There was no significant difference between both navigation methods. Discussion The number of navigated procedures in different fields is increasing due to promising results and a rising acceptance among surgeons. Computer and navigation technology is rapidly evolving thus making intraoperative navigation more efficient and user friendly. Commercial systems for neurosurgery and orthopaedic surgery have been developed and are evaluated at specialized centres. [9,10] These tools for image guided surgery have a high value particular in rigid tissues. Ultrasound based navigation As an alternative to intraoperative MRI scans 3D ultrasound has been evaluated for patients undergoing navigated brain tumor resection. It was shown that 3D ultrasound provides time- and cost effective 3D maps for neuronavigation with comparable results to MRI scans. [11] Reports on other clinical indications are limited and only few 3D ultrasound based systems have been developed and validated. [12] The problem of developing navigation techniques for abdominal and soft tissue tumors is mainly related to organ shift, tissue deformation and breathing artefacts. Different approaches to solve the difficulties have been evaluated. Birth et al. used an electromagnetic tracking system interlinked with a free hand 2D ultrasound probe to perform navigated biopsies and interventions (thermoablation). Their virtual needle guiding system facilitates an increased accuracy for needle placement with reduced trauma. Lesions difficult to reach and close to vulnerable structures were considered as potential clinical applications. [13] Related to the complexity of surgical procedures a transfer of navigation technologies from interventional procedures to the operation theatre is associated with additional difficulties. Different strategies have been evaluated. For example Marvik et al. developed a laparoscopic navigation pointer attached to an optoelectric tracking system, which provided an interactive 3D image based on preoperative CT/MRI scans during laparoscopic surgery. The additional information increased the intraoperative orientation through visualisation of vessels and tumors. [14] Augmenting preoperative image data with the intraoperative situation (registration) may lead to MDC Repository 2

4 significant inaccuracies especially in the absence of fixed anatomical landmarks. In an experimental study we were able to show that vessel based non-rigid registration of MRI/CT with intraoperative 3D ultrasound images is feasible. [15] Indirect navigation Addressing the various technological deficits we have developed a 3D ultrasound based navigation system for soft tissue tumor resection as described above. Two different techniques were evaluated in an experimental setting. During indirect navigation based on a guidance needle the surgeon benefited from the constant tracking of the instrument and the tumor, because the relative position of both the instrument and the tumor could be observed in real time on the navigation screen. The abstract visualisation of the tumor and the instruments was easy to understand and accurate resection could be supported by a virtual safety margin. The ultrasound guided introduction of the guidance needle may lead to positioning failures. In two cases the needle tip could not be placed exactly in the centre of the tumor resulting in a decreased accuracy of the resection margin. The virtual image of the tumor and the surrounding tissue may show discrepancies to the real anatomical situation resulting in further inaccuracies. Direct navigation During direct navigation based surgery ultrasound scans were repeated according to the intraoperative situation (e.g. manipulation). Usually between two to three intraoperative registrations were necessary resulting in a variable navigation time of min with a median of 19 min. Two ultrasound images were displayed in orthogonal planes (longitudinal and coronal) and showed sufficient information for spatial orientation. Advantages of this technique were the visualisation of anatomical landmarks (e.g. vessels) and a realistic depiction of the tumor boundaries. The additional anatomical visualisation was considered as potentially useful for safe and accurate resections in complex situations. The reacquisition of ultrasound images after tissue manipulation was necessary in all cases facilitating increased resection accuracy. The intermittent image acquisition may result in discordance. The median accuracy of direct and indirect navigation of 0.16 cm and 0.32 cm was significantly higher (p < 0.05) than that of conventional resection (0.42 cm). However, it must be emphasized that the tumor was poorly palpable. There was no significant difference between both navigation methods. Clinical application Our results demonstrate that 3D-ultrasound navigated resection of soft tissue tumors is feasible and may improve resection of non-palpable tumors such as small liver metastases. In a preliminary clinical study we showed that 3D-ultrasound based navigation for resection of liver metastases increased intraoperative orientation leading to higher accuracy and parenchyma preserving surgery. [16] Tracking systems A general disadvantage of using an optical tracking system is the necessity to provide a free line of sight between the camera and the tracked devices for an accurate registration process. To avoid signal interruptions we implemented an optical gate control which was visualized on the monitor. As an alternative navigation systems based on electromagnetic tracking techniques can be operated without a free line of sight. [13] The present generation of sensors is very small with improved accuracies of around 1mm. Interference with ferromagnetic materials (e.g. surgical instruments) remain a central disadvantage. Conclusion Compared to conventional surgery both navigation techniques allowed a more precise localisation of the tumor and a better control of the ideal resection margin. In the future a fusion of techniques may facilitate a continuous registration of a 3D ultrasound image. This could potentially improve the intraoperative handling and lead to an increased accuracy. These additional developments will be particularly necessary to adapt this navigation technique to clinical practice. We consider hepatic and soft tissue tumor resections as potential clinical applications. Conflicts of interest There is no financial and personal relationship with other people of organisations that could inappropriately bias this work. Corresponding Author S. Beller, Department of Surgery and Surgical Oncology, Campus Berlin Buch and Helios Hospital Berlin, Universitätsmedizin Berlin, Lindenberger Weg 80, Berlin, Germany. Tel.: ; fax: References 1. G. Unsgaard, S. Ommedal, T. Muller, A. Gronningsaeter and T.A. Nagelhus Hernes, Neuronavigation by intraoperative three-dimensional ultrasound: initial experience during brain tumor resection, Neurosurgery 50 (2002), pp A.J. Tria, The evolving role of navigation in minimally invasive total knee arthroplasty, Am J Orthop 35 (2006), pp W. Lamade, G. Glombitza and L. Fischer et al., The impact of 3-dimensional reconstructions on operation planning in liver surgery, Arch Surg 135 (11) (2000), pp A. Nabavi, P.M. Black and D.T. Gering et al., Serial intraoperative MR imaging of brain shift, Neurosurgery 48 (2001), pp C. Nimsky, O. Ganslandt and P. Hastreiter, Intraoperative compensation for brain shift, Surg Neurol 43 (4) (2001), pp M. Kleemann, P. Hildebrand, M. Birth and H.P. Bruch, Laparoscopic ultrasound navigation in liver surgery: MDC Repository 3

5 technical aspects and accuracy, Surg Endosc 20 (5) (2006), pp D.F. Leotta, P.R. Detmer and R.W. Martin, Performance of a miniature magnetic position sensor for three-dimensional ultrasound imaging, Ultrasound Med Biol 23 (4) (1997), pp P.R. Detmer, G. Bashein and T. Hodges, 3D ultrasonic image feature localization based on magnetic scanhead racking, Ultrasound Med Biol 20 (1994), pp D. Lindner, C. Trantakis and C. Renner et al., Application of intraoperative 3D ultrasound during navigated tumor resection, Minim Invasive Neurosurg 49 (4) (2006), pp K. Reijnders, M.H. Coppes, A.L.J. van Hulzen, J.P. Gravendeel, R.J. van Ginkel and H.J. Hoekstra, Image guided surgery: new technology for surgery of soft tissue and bone sarcomas, Eur J Surg Oncol 33 (3) (2007), pp O. Sergeeva, F. Uhlemann, G. Schackert, C. Hergeth, U. Morgenstern and R. Steinmeier, Integration of intraoperative 3D-ultrasound in a commercial navigation system, Zentralbl Neurochir 67 (4) (2006), pp J.H. Kaspersen, E. Soljie and J. Wesche et al., 3D ultrasound based navigation combined with preoperative CT during abdominal interventions: a feasability study, CardioVasc Intervent Radiol 26 (2003), pp M. Birth, P. Iblher, P. Hildebrand, J. Nolde and H.P. Bruch, Ultrasound-guided interventions using magnetic field navigation. First experiences with Ultra-Guide 2000 under operative conditions, Ultraschall Med 24 (2) (2003), pp R. Marvik, T. Lango and G.A. Tangen et al., Laparoscopic navigation pointer for three-dimensional image-guided surgery, Surg Endosc 18 (8) (2004), pp T. Lange, S. Eulenstein, M. Hunerbein and P.M. Schlag, Vessel-based non-rigid registration of MR/CT and 3D ultrasound for navigation in liver surgery, Comput Aided Surg 8 (5) (2003), pp S. Beller, M. Hünerbein, T. Lange, S. Eulenstein, B. Gebauer and P.M. Schlag, Image-guided surgery of liver metastases by three-dimensional ultrasound-based optoelectronic navigation, Br J Surg 94 (7) (2007), pp MDC Repository 4

6 Figure 1. Indirect navigation with an optically tracked electrocauter and needle device at the centre of the tumor (a). Instrument positions in relation to the tumor are displayed in real time in the coronal (b) and longitudinal plane (c). MDC Repository 5

7 Figure 2. Direct navigation with a tracked electrocauter and visualisation of two orthogonal ultrasound images. The white box displays intraoperative image registration with a tracked ultrasound probe (a). Ultrasound images of the silicon phantom in the coronal (b) and longitudinal plane (c) with the virtual scalpel. MDC Repository 6

8 Figure 3. CT scan of a resected specimen with the guidance needle inside the tumor after indirect navigation. The tumor centre is marked with a big red dot and the distance to the resection margin is measured at every 2 mm along the tumor surface (small red dots). MDC Repository 7

9 Figure 4. The deviation from the proposed security margin at 2 cm (red line) is shown as boxplots for each technique. The whiskers represent the minimum and maximum value, the upper and lower box borders show the 75% and 25% quartile and the black bar stands for the median. MDC Repository 8

10 Table 1: Statistical analysis (all values in centimetres) Conventional surgery (n=6) Indirect navigation (n=7) Direct navigation (n=10) Average minimal safety margin Absolute minimal safety margin Average maximal safety margin Absolute maximal safety margin Accuracy Standard deviation of the accuracy MDC Repository 9