- Email: [email protected]
The accuracy of image guided surgery based on cone beam computer tomography image data
The accuracy of image guided surgery based on cone beam computer tomography image data Georg Eggers, MD, DMD, PhD,a Hitomi Senoo, DDS, PhD,b Gavin Kane, MEng Sc,c and Joachim Mühling, MD, DMD, PhD,d Heidelberg, Germany, and Osaka, Japan HEIDELBERG UNIVERSITY HOSPITAL AND OSAKA UNIVERSITY GRADUATE SCHOOL OF DENTISTRY
Objective. The objective of this study was to verify if accurate patient-to-image registration for precision navigation in maxillofacial surgery is possible based on cone beam computed tomography (CBCT) image data. Study design. A maxillary registration template was placed on a standard plastic skull phantom that was equipped with a custom made model of the maxilla and with target markers. Imaging was performed with a CBCT device (Newtom 9000 Digital Volume Tomograph (DVT), QR s.r.l., Verona, Italy) and a computed tomography (CT) scanner (Somatom 4, Siemens, Forchheim, Germany). Using an infrared navigation system (Polaris, NDI, Waterloo, Ontario), multiple pair-point registration of both image data sets and the phantom were performed. The target registration error (TRE) was evaluated. Results. A total of 243 registrations were performed for either image data set. The spatial distribution of TRE on the skull showed increasing inaccuracy with growing distance from the registration markers. The average target registration error was 1.50 ⫾ 0.82 mm with CBCT and 1.57 ⫾ 0.84 mm with CT image data and did not differ significantly. Error distribution correlated strongly between CT- and CBCT-based registration. Conclusions. The overall registration accuracy based on CBCT image data was similar to CT. The strong correlation of the geometric distribution of TRE between CT- and CBCT-based measurements proves that CBCT can be equivalent to CT in image-guided maxillofacial surgery. (Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2009;107:e41-e48)
The development of image-guided surgery started in neurosurgery1 with the invention of stereotactic frame. The frame was attached to the patient’s head during image data acquisition and during the intervention, giving a coordinate system to both image space and patient space. The geometry of the frame provided the transform between either coordinate system. Hence it became possible to correlate positions in patient individual image data (e.g., x-ray,2 later also computed tomography [CT] and magnetic resonance imaging [MRI]3) accurately to positions in the patient’s head. While this concept was ideally suited for intracranial interventions, it proved to be cumbersome for facial surgery. With the development of frameless stereotaxy using navigation systems, this concept has evolved more and more in maxillofacial surgery.4 a
Senior Resident, Department of Oral and Cranio-Maxillofacial Surgery, Heidelberg University Hospital, Heidelberg, Germany. b Instructor, The First Department of Oral and Maxillofacial Surgery, Osaka University Graduate School of Dentistry, Osaka, Japan. c Researcher, Department of Oral and Cranio-Maxillofacial Surgery, Heidelberg University Hospital, Heidelberg, Germany. d Professor and Chairman, Department of Oral and Cranio-Maxillofacial Surgery, Heidelberg University Hospital, Heidelberg, Germany. Received for publication May 25, 2008; returned for revision Oct 3, 2008; accepted for publication Oct 31, 2008. 1079-2104/$ - see front matter © 2009 Mosby, Inc. All rights reserved. doi:10.1016/j.tripleo.2008.10.022
Navigation systems allow the surgeon to choose an individual configuration of registration markers. These markers are identified in image data using a software tool, and on the patient using a tracked pointing device. This information provides the transform between image space and patient space. It is obvious that the geometric accuracy and homogeneity of the image data are of paramount importance for the accuracy of this transform and hence for the accurate guidance of the surgeon. The imaging modality predominantly used in imageguided maxillofacial surgery is CT.5 In past years, cone beam computed tomography (CBCT) has become more and more available as an alternative to CT for imaging in the region of the head. The main advantage is that there is a considerably lower exposure to x-rays than in CT.6 Image quality of CBCT is not as good as in dental CT, which results in poorer imaging of enamel-dentin interface, pulp cavity’s edges and bony structures surrounding the teeth, and poor or no accessibility of the periodontal ligament space.7 However, CBCT has fewer metal artifacts than CT7 and is successful in clinical use, e.g., in patients with cleft lip and palate,8 orthodontic treatment,9 or mandibular osteomyelitis.10 Articles about the use of CBCT for surgical navigation mainly report the use in dental implant placement.11 There are only sporadic reports about the use of CBCT image data as the basis for image-guided maxillofacial surgery.12 Routinely, CT is used at our institution for imagee41
e42
Eggers et al.
OOOOE March 2009
Fig. 1. Registration phantom made from a resin skull, equipped with titanium screws as target markers (top left). Frontal view (top right), top view (bottom left) and side view (bottom right) of a transparent reconstruction of the phantom (gray) with the positions of the marker screws (white) on skull phantom and registration template.
guided surgery. Maxillary mounted templates with fiducial markers for patient-to-image registration are used. This method has proven to be reliable and accurate.13 A prerequisite for accurate registration is a high geometric accuracy of the image data.4 However, studies on geometric accuracy of CBCT devices report considerable differences between various devices.14 At our institution, a CBCT device (Newtom 9000 Digital Volume Tomograph (DVT), QR s.r.l., Verona, Italy) is in routine use in patient care. Its geometric accuracy is slightly lower than that of a multislice CT scanner.15 Hence, the objective of this study was to evaluate whether the image data from this CBCT device allows accurate patient-to-image registration and navigation. Is CBCT-based navigation possible as accurately as with CT image data, despite the slightly lower geometric accuracy? MATERIALS AND METHODS Phantom The registration accuracy of CBCT image data was evaluated in a phantom study. A registration phantom was built using a resin skull (3B Scientific, Dresden, Germany). The maxilla of this plastic skull represented bone and teeth only. To provide a more realistic interface for the registration template, the maxilla was cut off and was replaced with a model of the maxilla with gingiva, mucosa, and complete dentition. This model was created from an impression using a resin (ExaktoForm, Bredent, Senden, Germany) and was glued rigidly to the skull phantom (Fig. 1). Sixty-two titanium screws (Stryker Leibinger Micro Implants, Freiburg, Germany) were attached externally
Fig. 2. A, Maxillary registration templates for image-guided oral and maxillofacial surgery from 3 patient cases: the 5 titanium screw fiducial markers are distributed over the arch of the dentition. The screw heads are used as fiducial markers for patient to image registration. The geometry of the fiducial markers is variable. B, Maxillary registration template as used in the study: the titanium screw fiducial markers are placed in 5 groups of 3 markers over the arch of the dentition.
and internally to the skull (Fig. 1). They were used as target markers for measurement of targeting accuracy using the navigation system. Registration template A standard method for patient-to-image registration in maxillofacial surgery is pair-point registration using a maxillary template equipped with radio-opaque fiducial markers. At our institution, the standard configuration for clinical routine consists of a maxillary template made of a light-curing resin (Triad Gel, Dentsply, York, PA). It is produced on a plaster model from an alginate impression of the maxilla. As fiducial markers, 5 titanium screws (Stryker Leibinger Micro Implants, Freiburg, Germany) were used. They were attached rigidly to the template using the same light-curing resin and were evenly distributed over the arch of the template; however, the position of these fiducial markers was always slightly different (Fig. 2). This is because of the variability in the geometry of the maxilla and be-
OOOOE Volume 107, Number 3
Eggers et al. e43
cause the markers are placed free-handedly by the dental technician (Fig. 2, A). Hence, to reflect this variability in marker configuration in the experimental setup, the design of the registration template was modified for this study. In each of the 5 regions on the arch, 3 markers were placed instead of 1 (Fig. 2, B). For patient-to-image registration experiments, repeated registrations were performed with various combinations of the fiducial markers to reflect this variability in fiducial marker geometry. Imaging CBCT and CT image data of the phantom with the attached maxillary registration template were acquired. Spiral CT image data were acquired using a CT scanner (Somatom, Siemens, Forchheim, Germany). The field of view was 219.6 by 219.6 by 168.0 mm (Fig. 3, A). Tube voltage was 120 kV, tube current 60 mA, pitch 1. The in-plane size of image pixels was 0.43 ⫻ 0.43 mm. Image data were reconstructed with a slice thickness of 1 mm. Axial CBCT imaging was performed with a Newtom 9000 DVT (QR s.r.l., Verona, Italy). The field of view was cylindrical with a diameter of 148.19 mm and a height of 96.00 mm (Fig. 3, B). Tube voltage was 110 kV; tube current was set automatically by the device. The image data were reconstructed with a slice thickness of 1 mm. The in-plane size of image pixels was 0.29 ⫻ 0.29 mm. Registration All registration experiments were performed independently and in identical manner on the CT and the CBCT image data. In preregistration, any combination of 5 marker coordinates in image data was used; 1 of 3 markers at each of the 5 locations was selected. From the image data, the positions of the fiducial markers and target markers were recorded in triplanar (axial, coronal, sagittal) 2-dimensional views of the CT or CBCT data set respectively. The template was attached to the maxilla and a tracking body carrying infrared light-emitting diodes (IR-LED) was attached (Fig. 4). Measurements of fiducial and target marker positions in physical space of the phantom were performed using an infrared tracking system that is used in many commercially available surgical navigation systems (Polaris, NDI, Waterloo, Ontario, Canada). Using a tracked pointing device, the positions of the fiducial markers on the template were recorded. Fiducial marker coordinates from physical space and image space were aligned using a least-squares rigid body transformation.16 This registration transform was
Fig. 3. A, Visualization of the CT scanning volume: the blue cube indicates the border of the volume of interest that was imaged using the Siemens Somatom CT. Within the volume is the complete skull phantom (white) with all fiducial markers and all target markers (red). B, Visualization of the CBCT scanning volume: the blue cylindroid shape indicates the border of the volume of interest that was imaged using the Newtom 9000 DVT. Within the volume are parts of the skull phantom (white) with all fiducial markers and a subset of the target markers (red).
then applied to the coordinates of the target markers on the skull that were also measured in physical space using the navigation system’s tracked pointer (Fig. 5). Registrations were calculated with all possible fiducial marker combinations. Hence, a total of 243 (3 to
e44
Eggers et al.
Fig. 4. Maxillary template (A) attached to the phantom (B) for the registration experiments in this study. The titanium screws are placed in 5 groups of 3 to reflect the variability of the position of the registration markers on the dental arch. The photograph shows the infrared tracking body (C) that was attached to the template after imaging for the registration experiments.
the power of 5) target registrations based on the varying configurations of the fiducial markers were performed. The Euclidean distance of these target coordinates after registration to their counterparts in image space was the target registration error (TRE). It was the measure for navigation accuracy. For this study a software tool was created using C⫹⫹ and the visualization toolkit (vtk).17 This software was used to read the coordinates of fiducial and target marker positions in image space and from the infrared tracking system, perform the transformation calculations and calculate TRE values. All visualizations were created with ParaView 2.6 (Kitware, Clifton Park, NY). TRE values from CT- and CBCT-based navigation were compared using paired 2-tailed t tests. The correlations between both data sets were calculated using Pearsons test.
OOOOE March 2009
Fig. 5. Flowchart for the experimental measurement of navigation accuracy.
RESULTS The CT scan covered the whole skull. All 15 registration markers and all 62 target markers were within the imaged volume (Fig. 3, A). In the CBCT image data set, all 15 registration markers of the maxillary splint were within the imaged volume, but only 35 of the 62 target markers on the phantom skull. This was because of the restricted volume of interest of the CBCT scanner that would not cover the whole skull phantom (Fig. 3, B). Evaluation of navigation accuracy was performed using only those 35 target markers that were also visible in both CT and CBCT. The registration markers on the maxillary splint could be identified easily in CT and CBCT image data. The representation of the markers in this phantom study was similar to CT or CBCT imaging of registration templates in patient cases (Fig. 6). There were 243 registrations performed for each image data set and the TRE for all 35 target markers in each registration was calculated for CT- and CBCT-based navigation.
OOOOE Volume 107, Number 3
Fig. 6. Imaging of the registration splint with the titanium screw markers (arrows) in phantom study and patient cases. A, Axial CBCT image of the phantom. B, Axial CBCT image of a patient case. C, Axial CT image of the phantom. D, Axial CT image of a patient case.
The average TRE of the 35 markers in CBCT image data was 1.50 ⫾ 0.82 mm (mean ⫾ SD) (Fig. 7). The average TRE of those 35 markers in CT image data was 1.57 ⫾ 0.84 mm (mean ⫾ SD) (Fig. 8). This difference in TRE between CT- and CBCT-based registration was not significant (P ⫽ .15, paired 2-tailed t test) (Fig. 9). TRE at the target markers was higher with increasing distance from the registration markers. The average TRE of each target in CBCT-based registration correlated significantly with the average TRE in CT-based registration (Pearsons r ⫽ 0.94, P ⬍ .0001) (Fig. 10). DISCUSSION The accuracy of image-guided surgery depends on the geometry of the fiducial marker configuration and its relation to the surgical target. Furthermore, navigation accuracy depends on the accuracy of fiducial marker localization. Hence, geometric accuracy of the image data is a prerequisite for accurate navigation. The registration method used in this study is in routine use at our institution and has proven its accuracy and reliability in image-guided surgery of the viscerocranium and the anterior skull base based on CT image data.13,18 Accuracy measurements exist for CBCT-based navigation in the oral cavity, close to the fiducial registration markers. Extraoral CBCT-guided navigation was performed for the retrieval of foreign bodies12 or orbital reconstruction.19 However, no evaluations of the accu-
Eggers et al. e45
Fig. 7. Distribution of the average target registration error in CBCT imaging. Because of the restricted scanning volume, the phantom skull is not imaged completely. The colored spheres indicate the locations of the target markers. The color indicates the average TRE for all registrations at the respective target marker.
Fig. 8. Distribution of the average target registration error (TRE) in CT imaging. The colored spheres indicate the locations of the target markers. The color indicates the average TRE. Small white spheres indicate target markers that were not evaluated because they were not visible in the corresponding CBCT image data set.
racy exist. Since the distance to the registration markers is larger, higher inaccuracies are to be expected.16 Hence, it was the objective of this study to determine the accuracy that can be achieved for CBCT-based navigation in midface and skull base and compared to the gold standard CT. Previous measurements indicated that the geometric accuracy of the Newtom 9000 DVT is slightly lower than that of CT, particularly in the marginal regions of the imaging volume. This is in accordance with pub-
e46
Eggers et al.
Fig. 9. Target registration errors at the various marker positions in CT and CBCT image data. Each dot indicates the TRE at one marker. The horizontal lines indicate the average TRE for CT and CBCT image data, respectively.
Fig. 10. Correlation analysis of registration accuracy. Each ⫹ indicates the TRE in CT (x-axis) and CBCT (y-axis) of a target marker position.
lished reports of higher inaccuracy of geometric measurements in marginal regions of CBCT image data.20 With the registration experiments we wanted to specify the resulting targeting accuracy when using CBCT image data. Since the imaging volume of the used device is limited, it is most likely that registration markers are located at the margin of the volume imaged. Hence the image data used for this study is a
OOOOE March 2009
less-than-optimum, though realistic configuration, for the position of the registration markers in image data for image guided maxillofacial surgery. The results of the registration experiments showed that there was no significant difference in the targeting accuracy when CT or CBCT image data were used. The average TRE values, which indicate the accuracy with which the surgeon is guided by the navigation system, were similar. Furthermore, when comparing the imaging modalities, there was a high correlation of TRE for every single target in CT and CBCT. This indicates that the surgeon can expect to have the same navigation accuracy based on CBCT image data that he or she would have had using CT instead. Hence it can be concluded, that the observed irregularities in the geometric accuracy of CBCT image data do not affect the registration accuracy. This can be explained by the fact that there are many more sources of error in imageguided surgery.21 Contributing to the overall error in navigation accuracy are the relocation error of the registration template, which is usually temporarily removed between imaging and surgery,22 inaccuracies of the tracking system,23 erroneous identification of the fiducial markers in image data and in patient space, and the geometric configuration of the registration markers and the position of a target relatively to the registration markers.24 From the technical point of view the use of CBCT image data for navigation is a good alternative to CT. Besides similar registration accuracy, there is a lower radiation dose6 and fewer dental metal artifacts.7 A general problem of phantom studies is the validity of the results on a real patient. The key issue in this study are the geometric distribution and identification of markers in a plastic phantom skull setting. Previous studies showed consistently good geometric accuracy of CBCT image data in plastic phantoms as well as on human.25,26 The slight decay of Newtom CBCT geometric accuracy to the margins of the volume of interest was found in plastic phantom studies as well as in human skull studies.15,20 Furthermore, we found a good representation of the registration markers in the phantom CT and CBCT image data, similar to clinical CT and CBCT image data (Fig. 6). Hence, we can assume that the use of a plastic phantom did not compromise the results. However, a problem that could not be addressed in this study was the problem of motion artifacts. There is always some motion of the patient during image data.27 This can reduce navigation accuracy.28 However, there are possible clinical circumstances that might restrict the use of CBCT. First, situations where high soft tissue contrast or the use of contrast enhancement agents is required. Here CT has better
OOOOE Volume 107, Number 3
diagnostic image quality. Another restriction might be the limited field of view of CBCT scanners in comparison to CT. The current development of CBCT imaging hardware shows 2 trends. First, by using larger detector units (e.g., 12 inches diameter instead of 9 inches in this study), imaging systems with larger or variable field of view become available.6 This mitigates the problem of restricted field of view for navigation purposes. Second, there is a change in detector technology. Flat-panel detectors (FPD) became more and more widespread as an alternative to the image intensifier (II) technology used in this study.29,30 Comparative studies in a variety of settings have shown that FPD devices can provide image data with less noise,31 fewer artifacts,32 lower radiation dose,33 and higher spatial resolution.34 However, FPD technology is still developing. Experience from cardiac imaging shows that FPD systems can be, but not necessarily are, better than II systems. Other components like x-ray tube, generator, and dose control contribute to image quality.35 There exists no definition of minimum navigation accuracy in literature. For paranasal sinus and anterior skull base surgery system accuracies of 1 to 2 mm were labeled “acceptable.”36 In a comparison of 5 navigation systems, accuracies of 1.76 to 3.22 mm were found to have “reached the high standard of precision necessary for interventions in clinical use.”37 An accuracy of up to 2 mm increased safety for patient and surgeon in lateral skull base surgery.38 The accuracy values found in this study show the equivalent accuracy of CBCT-based navigation as compared to the gold standard CT-based navigation. Whether the accuracies are sufficient or not depends on the individual surgical task. The results may at best support the surgeon in this decision by providing concrete numbers on the expected inaccuracy. CONCLUSION Registration of CBCT image data to the patient’s body for image-guided surgery is possible with similar geometric accuracy as that of CT. Hence, this imaging modality is appropriate as a basis for image-guided surgery of the face and anterior skull base. REFERENCES 1. Horsley VA, Clarke RH. The structure and functions of the cerebellum examined by a new method. Brain 1908;31:45-124. 2. Al-Rodhan NR, Kelly PJ. Pioneers of stereotactic neurosurgery. Stereotact Funct Neurosurg 1992;58:60-6. 3. Lunsford LD, Martinez AJ, Latchaw RE. Stereotaxic surgery with a magnetic resonance- and computerized tomography-compatible system. J Neurosurg 1986;64:872-8. 4. Eggers G, Mühling J, Marmulla R. Image-to-patient registration techniques in head surgery. Int J Oral Maxillofac Surg 2006;35: 1081-95.
Eggers et al. e47 5. Mischkowski RA, Zinser MJ, Ritter L, Neugebauer J, Keeve E, Zöller JE. Intraoperative navigation in the maxillofacial area based on 3D imaging obtained by a cone-beam device. Int J Oral Maxillofac Surg 2007;36:687-94. 6. Ludlow JB, Davies-Ludlow LE, Brooks SL, Howerton WB. Dosimetry of 3 CBCT devices for oral and maxillofacial radiology: CB Mercuray, NewTom 3G and i-CAT. Dentomaxillofac Radiol 2006;35:219-26. 7. Holberg C, Steinhäuser S, Geis P, Rudzki-Janson I. Cone-beam computed tomography in orthodontics: benefits and limitations. J Orofac Orthop 2005;66:434-44. 8. Wörtche R, Hassfeld S, Lux CJ, Müssig E, Hensley FW, Krempien R, et al. Clinical application of cone beam digital volume tomography in children with cleft lip and palate. Dentomaxillofac Radiol 2006;35:88-94. 9. Müssig E, Wörtche R, Lux CJ. Indications for digital volume tomography in orthodontics. J Orofac Orthop 2005;66:241-9. 10. Schulze D, Blessmann M, Pohlenz P, Wagner KW, Heiland M. Diagnostic criteria for the detection of mandibular osteomyelitis using cone-beam computed tomography. Dentomaxillofac Radiol 2006;35:232-5. 11. Van Assche N, van Steenberghe D, Guerrero ME, Hirsch E, Schutyser F, Quirynen M, et al. Accuracy of implant placement based on pre-surgical planning of three-dimensional cone-beam images: a pilot study. J Clin Periodontol 2007;34:816-21. 12. Pohlenz P, Blessmann M, Blake F, Heinrich S, Schmelzle R, Heiland M. Clinical indications and perspectives for intraoperative cone-beam computed tomography in oral and maxillofacial surgery. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2007;103:412-7. 13. Eggers G, Mühling J, Marmulla R. Template-based registration for image-guided maxillofacial surgery. J Oral Maxillofac Surg 2005;63:1330-6. 14. Stratemann SA, Huang JC, Maki K, Miller AJ, Hatcher DC. Comparison of cone beam computed tomography imaging with physical measures. Dentomaxillofac Radiol 2008;37:80-93. 15. Eggers G, Klein J, Welzel T, Mühling J. Geometric accuracy of digital volume tomography and conventional computed tomography. Br J Oral Maxillofac Surg 2008;46:639-44. 16. West JB, Fitzpatrick JM, Toms SA, Maurer CR Jr, Maciunas RJ. Fiducial point placement and the accuracy of point-based, rigid body registration. Neurosurgery 2001;48:810-6. 17. Schroeder W, Martin K, Lorensen W. The visualization toolkit, an object-oriented approach to 3D graphics. 3rd ed. Clifton Park, NY: Kitware Inc.; 2003. 18. Eggers G, Mühling J. Template-based registration for imageguided skull base surgery. Otolaryngol Head Neck Surg 2007;136:907-13. 19. Zizelmann C, Gellrich NC, Metzger MC, Schön R, Schmelzeisen R, Schramm A. Computer-assisted reconstruction of orbital floor based on cone beam tomography. Br J Oral Maxillofac Surg 2007;45:79-80. 20. Lascala CA, Panella J, Marques MM. Analysis of the accuracy of linear measurements obtained by cone beam computed tomography (CBCT-NewTom). Dentomaxillofac Radiol 2004;33: 291-4. 21. Ecke U, Maurer J, Boor S, Khan M, Mann WJ. Fehlerquellen der Navigation in der lateralen Schädelbasischirurgie. Darstellung von Einflussfaktoren in der Praxis. HNO 2003;51:386-93. 22. Fitzpatrick JM, Balachandran R, Labadie RF. Bite-block relocation error in image-guided otologic surgery. In: Barillot C, Haynor DR, Hellier P, editors. MICCAI 2004, Lect Notes Comput Sci 3217. Berlin Heidelberg: Springer; 2004. p. 518-25. 23. Morris A. Bone registration and tracking using an optical track-
e48
24.
25.
26.
27.
28.
29.
30.
31.
32.
OOOOE March 2009
Eggers et al.
ing system [master’s thesis]. Baltimore: Johns Hopkins University; 2001. Fitzpatrick JM, West JB, Maurer CR Jr. Predicting error in rigid-body point-based registration. IEEE Trans Med Imaging 1998;17:694-702. Ludlow JB, Laster WS, See M, Bailey LJ, Hershey HG. Accuracy of measurements of mandibular anatomy in cone beam computed tomography images. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2007;103:534-42. Stratemann SA, Huang JC, Maki K, Miller AJ, Hatcher DC. Comparison of cone beam computed tomography imaging with physical measures. Dentomaxillofac Radiol 2008;37:80-93. Bontempi M, Bettuzzi M, Casali F, Pasini A, Rossi A, Ariu M. Relevance of head motion in dental cone-beam CT scanner images depending on patient positioning. Int J CARS 2008; 3:249-55. Marmulla R, Mühling J. The influence of computed tomography motion artifacts on computer-assisted surgery. J Oral Maxillofac Surg 2006;64:466-70. Honey OB, Scarfe WC, Hilgers MJ, Klueber K, Silveira AM, Haskell BS, et al. Accuracy of cone-beam computed tomography imaging of the temporomandibular joint: comparisons with panoramic radiology and linear tomography. Am J Orthod Dentofacial Orthop 2007;132:429-38. Guerrero ME, Jacobs R, Loubele M, Schutyser F, Suetens P, van Steenberghe D. State-of-the-art on cone beam CT imaging for preoperative planning of implant placement. Clin Oral Investig 2006;10:1-7. Baba R, Ueda K, Okabe M. Using a flat-panel detector in high resolution cone beam CT for dental imaging. Dentomaxillofac Radiol 2004;33:285-90. Katsumata A, Hirukawa A, Okumura S, Naitoh M, Fujishita M,
33.
34.
35.
36.
37.
38.
Ariji E, et al. Effects of image artifacts on gray-value density in limited-volume cone-beam computerized tomography. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2007;104:829-36. Hatakeyama Y, Kakeda S, Ohnari N, Moriya J, Oda N, Nishino K, et al. Reduction of radiation dose for cerebral angiography using flat panel detector of direct conversion type: a vascular phantom study. AJNR Am J Neuroradiol 2007;28:645-50. Baba R, Konno Y, Ueda K, Ikeda S. Comparison of flat-panel detector and image-intensifier detector for cone-beam CT. Comput Med Imaging Graph 2002;26:153-8. Davies AG, Cowen AR, Kengyelics SM, Moore J, Sivananthan MU. Do flat detector cardiac x-ray systems convey advantages over image-intensifier-based systems? Study comparing x-ray dose and image quality. Eur Radiol 2007;17:1787-94. Cartellieri M, Vorbeck F. Endoscopic sinus surgery using intraoperative computed tomography imaging for updating a threedimensional navigation system. Laryngoscope 2000;110:292-6. Cartellieri M, Kremser J, Vorbeck F. Comparison of different 3D navigation systems by a clinical “user.” Eur Arch Otorhinolaryngol 2001;258:38-41. Caversaccio M, Bächler R, Lädrach K, Schroth G, Nolte LP, Häusler R. The “Bernese” Frameless optical computer aided surgery system. Comp Aid Surg 1999;4:328-34.
Reprint requests: Georg Eggers, MD, DMD Department of Oral and Cranio-Maxillofacial Surgery Heidelberg University Hospital Im Neuenheimer Feld 400 69120 Heidelberg, Germany [email protected]
Objective. The objective of this study was to verify if accurate patient-to-image registration for precision navigation in maxillofacial surgery is possible based on cone beam computed tomography (CBCT) image data. Study design. A maxillary registration template was placed on a standard plastic skull phantom that was equipped with a custom made model of the maxilla and with target markers. Imaging was performed with a CBCT device (Newtom 9000 Digital Volume Tomograph (DVT), QR s.r.l., Verona, Italy) and a computed tomography (CT) scanner (Somatom 4, Siemens, Forchheim, Germany). Using an infrared navigation system (Polaris, NDI, Waterloo, Ontario), multiple pair-point registration of both image data sets and the phantom were performed. The target registration error (TRE) was evaluated. Results. A total of 243 registrations were performed for either image data set. The spatial distribution of TRE on the skull showed increasing inaccuracy with growing distance from the registration markers. The average target registration error was 1.50 ⫾ 0.82 mm with CBCT and 1.57 ⫾ 0.84 mm with CT image data and did not differ significantly. Error distribution correlated strongly between CT- and CBCT-based registration. Conclusions. The overall registration accuracy based on CBCT image data was similar to CT. The strong correlation of the geometric distribution of TRE between CT- and CBCT-based measurements proves that CBCT can be equivalent to CT in image-guided maxillofacial surgery. (Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2009;107:e41-e48)
The development of image-guided surgery started in neurosurgery1 with the invention of stereotactic frame. The frame was attached to the patient’s head during image data acquisition and during the intervention, giving a coordinate system to both image space and patient space. The geometry of the frame provided the transform between either coordinate system. Hence it became possible to correlate positions in patient individual image data (e.g., x-ray,2 later also computed tomography [CT] and magnetic resonance imaging [MRI]3) accurately to positions in the patient’s head. While this concept was ideally suited for intracranial interventions, it proved to be cumbersome for facial surgery. With the development of frameless stereotaxy using navigation systems, this concept has evolved more and more in maxillofacial surgery.4 a
Senior Resident, Department of Oral and Cranio-Maxillofacial Surgery, Heidelberg University Hospital, Heidelberg, Germany. b Instructor, The First Department of Oral and Maxillofacial Surgery, Osaka University Graduate School of Dentistry, Osaka, Japan. c Researcher, Department of Oral and Cranio-Maxillofacial Surgery, Heidelberg University Hospital, Heidelberg, Germany. d Professor and Chairman, Department of Oral and Cranio-Maxillofacial Surgery, Heidelberg University Hospital, Heidelberg, Germany. Received for publication May 25, 2008; returned for revision Oct 3, 2008; accepted for publication Oct 31, 2008. 1079-2104/$ - see front matter © 2009 Mosby, Inc. All rights reserved. doi:10.1016/j.tripleo.2008.10.022
Navigation systems allow the surgeon to choose an individual configuration of registration markers. These markers are identified in image data using a software tool, and on the patient using a tracked pointing device. This information provides the transform between image space and patient space. It is obvious that the geometric accuracy and homogeneity of the image data are of paramount importance for the accuracy of this transform and hence for the accurate guidance of the surgeon. The imaging modality predominantly used in imageguided maxillofacial surgery is CT.5 In past years, cone beam computed tomography (CBCT) has become more and more available as an alternative to CT for imaging in the region of the head. The main advantage is that there is a considerably lower exposure to x-rays than in CT.6 Image quality of CBCT is not as good as in dental CT, which results in poorer imaging of enamel-dentin interface, pulp cavity’s edges and bony structures surrounding the teeth, and poor or no accessibility of the periodontal ligament space.7 However, CBCT has fewer metal artifacts than CT7 and is successful in clinical use, e.g., in patients with cleft lip and palate,8 orthodontic treatment,9 or mandibular osteomyelitis.10 Articles about the use of CBCT for surgical navigation mainly report the use in dental implant placement.11 There are only sporadic reports about the use of CBCT image data as the basis for image-guided maxillofacial surgery.12 Routinely, CT is used at our institution for imagee41
e42
Eggers et al.
OOOOE March 2009
Fig. 1. Registration phantom made from a resin skull, equipped with titanium screws as target markers (top left). Frontal view (top right), top view (bottom left) and side view (bottom right) of a transparent reconstruction of the phantom (gray) with the positions of the marker screws (white) on skull phantom and registration template.
guided surgery. Maxillary mounted templates with fiducial markers for patient-to-image registration are used. This method has proven to be reliable and accurate.13 A prerequisite for accurate registration is a high geometric accuracy of the image data.4 However, studies on geometric accuracy of CBCT devices report considerable differences between various devices.14 At our institution, a CBCT device (Newtom 9000 Digital Volume Tomograph (DVT), QR s.r.l., Verona, Italy) is in routine use in patient care. Its geometric accuracy is slightly lower than that of a multislice CT scanner.15 Hence, the objective of this study was to evaluate whether the image data from this CBCT device allows accurate patient-to-image registration and navigation. Is CBCT-based navigation possible as accurately as with CT image data, despite the slightly lower geometric accuracy? MATERIALS AND METHODS Phantom The registration accuracy of CBCT image data was evaluated in a phantom study. A registration phantom was built using a resin skull (3B Scientific, Dresden, Germany). The maxilla of this plastic skull represented bone and teeth only. To provide a more realistic interface for the registration template, the maxilla was cut off and was replaced with a model of the maxilla with gingiva, mucosa, and complete dentition. This model was created from an impression using a resin (ExaktoForm, Bredent, Senden, Germany) and was glued rigidly to the skull phantom (Fig. 1). Sixty-two titanium screws (Stryker Leibinger Micro Implants, Freiburg, Germany) were attached externally
Fig. 2. A, Maxillary registration templates for image-guided oral and maxillofacial surgery from 3 patient cases: the 5 titanium screw fiducial markers are distributed over the arch of the dentition. The screw heads are used as fiducial markers for patient to image registration. The geometry of the fiducial markers is variable. B, Maxillary registration template as used in the study: the titanium screw fiducial markers are placed in 5 groups of 3 markers over the arch of the dentition.
and internally to the skull (Fig. 1). They were used as target markers for measurement of targeting accuracy using the navigation system. Registration template A standard method for patient-to-image registration in maxillofacial surgery is pair-point registration using a maxillary template equipped with radio-opaque fiducial markers. At our institution, the standard configuration for clinical routine consists of a maxillary template made of a light-curing resin (Triad Gel, Dentsply, York, PA). It is produced on a plaster model from an alginate impression of the maxilla. As fiducial markers, 5 titanium screws (Stryker Leibinger Micro Implants, Freiburg, Germany) were used. They were attached rigidly to the template using the same light-curing resin and were evenly distributed over the arch of the template; however, the position of these fiducial markers was always slightly different (Fig. 2). This is because of the variability in the geometry of the maxilla and be-
OOOOE Volume 107, Number 3
Eggers et al. e43
cause the markers are placed free-handedly by the dental technician (Fig. 2, A). Hence, to reflect this variability in marker configuration in the experimental setup, the design of the registration template was modified for this study. In each of the 5 regions on the arch, 3 markers were placed instead of 1 (Fig. 2, B). For patient-to-image registration experiments, repeated registrations were performed with various combinations of the fiducial markers to reflect this variability in fiducial marker geometry. Imaging CBCT and CT image data of the phantom with the attached maxillary registration template were acquired. Spiral CT image data were acquired using a CT scanner (Somatom, Siemens, Forchheim, Germany). The field of view was 219.6 by 219.6 by 168.0 mm (Fig. 3, A). Tube voltage was 120 kV, tube current 60 mA, pitch 1. The in-plane size of image pixels was 0.43 ⫻ 0.43 mm. Image data were reconstructed with a slice thickness of 1 mm. Axial CBCT imaging was performed with a Newtom 9000 DVT (QR s.r.l., Verona, Italy). The field of view was cylindrical with a diameter of 148.19 mm and a height of 96.00 mm (Fig. 3, B). Tube voltage was 110 kV; tube current was set automatically by the device. The image data were reconstructed with a slice thickness of 1 mm. The in-plane size of image pixels was 0.29 ⫻ 0.29 mm. Registration All registration experiments were performed independently and in identical manner on the CT and the CBCT image data. In preregistration, any combination of 5 marker coordinates in image data was used; 1 of 3 markers at each of the 5 locations was selected. From the image data, the positions of the fiducial markers and target markers were recorded in triplanar (axial, coronal, sagittal) 2-dimensional views of the CT or CBCT data set respectively. The template was attached to the maxilla and a tracking body carrying infrared light-emitting diodes (IR-LED) was attached (Fig. 4). Measurements of fiducial and target marker positions in physical space of the phantom were performed using an infrared tracking system that is used in many commercially available surgical navigation systems (Polaris, NDI, Waterloo, Ontario, Canada). Using a tracked pointing device, the positions of the fiducial markers on the template were recorded. Fiducial marker coordinates from physical space and image space were aligned using a least-squares rigid body transformation.16 This registration transform was
Fig. 3. A, Visualization of the CT scanning volume: the blue cube indicates the border of the volume of interest that was imaged using the Siemens Somatom CT. Within the volume is the complete skull phantom (white) with all fiducial markers and all target markers (red). B, Visualization of the CBCT scanning volume: the blue cylindroid shape indicates the border of the volume of interest that was imaged using the Newtom 9000 DVT. Within the volume are parts of the skull phantom (white) with all fiducial markers and a subset of the target markers (red).
then applied to the coordinates of the target markers on the skull that were also measured in physical space using the navigation system’s tracked pointer (Fig. 5). Registrations were calculated with all possible fiducial marker combinations. Hence, a total of 243 (3 to
e44
Eggers et al.
Fig. 4. Maxillary template (A) attached to the phantom (B) for the registration experiments in this study. The titanium screws are placed in 5 groups of 3 to reflect the variability of the position of the registration markers on the dental arch. The photograph shows the infrared tracking body (C) that was attached to the template after imaging for the registration experiments.
the power of 5) target registrations based on the varying configurations of the fiducial markers were performed. The Euclidean distance of these target coordinates after registration to their counterparts in image space was the target registration error (TRE). It was the measure for navigation accuracy. For this study a software tool was created using C⫹⫹ and the visualization toolkit (vtk).17 This software was used to read the coordinates of fiducial and target marker positions in image space and from the infrared tracking system, perform the transformation calculations and calculate TRE values. All visualizations were created with ParaView 2.6 (Kitware, Clifton Park, NY). TRE values from CT- and CBCT-based navigation were compared using paired 2-tailed t tests. The correlations between both data sets were calculated using Pearsons test.
OOOOE March 2009
Fig. 5. Flowchart for the experimental measurement of navigation accuracy.
RESULTS The CT scan covered the whole skull. All 15 registration markers and all 62 target markers were within the imaged volume (Fig. 3, A). In the CBCT image data set, all 15 registration markers of the maxillary splint were within the imaged volume, but only 35 of the 62 target markers on the phantom skull. This was because of the restricted volume of interest of the CBCT scanner that would not cover the whole skull phantom (Fig. 3, B). Evaluation of navigation accuracy was performed using only those 35 target markers that were also visible in both CT and CBCT. The registration markers on the maxillary splint could be identified easily in CT and CBCT image data. The representation of the markers in this phantom study was similar to CT or CBCT imaging of registration templates in patient cases (Fig. 6). There were 243 registrations performed for each image data set and the TRE for all 35 target markers in each registration was calculated for CT- and CBCT-based navigation.
OOOOE Volume 107, Number 3
Fig. 6. Imaging of the registration splint with the titanium screw markers (arrows) in phantom study and patient cases. A, Axial CBCT image of the phantom. B, Axial CBCT image of a patient case. C, Axial CT image of the phantom. D, Axial CT image of a patient case.
The average TRE of the 35 markers in CBCT image data was 1.50 ⫾ 0.82 mm (mean ⫾ SD) (Fig. 7). The average TRE of those 35 markers in CT image data was 1.57 ⫾ 0.84 mm (mean ⫾ SD) (Fig. 8). This difference in TRE between CT- and CBCT-based registration was not significant (P ⫽ .15, paired 2-tailed t test) (Fig. 9). TRE at the target markers was higher with increasing distance from the registration markers. The average TRE of each target in CBCT-based registration correlated significantly with the average TRE in CT-based registration (Pearsons r ⫽ 0.94, P ⬍ .0001) (Fig. 10). DISCUSSION The accuracy of image-guided surgery depends on the geometry of the fiducial marker configuration and its relation to the surgical target. Furthermore, navigation accuracy depends on the accuracy of fiducial marker localization. Hence, geometric accuracy of the image data is a prerequisite for accurate navigation. The registration method used in this study is in routine use at our institution and has proven its accuracy and reliability in image-guided surgery of the viscerocranium and the anterior skull base based on CT image data.13,18 Accuracy measurements exist for CBCT-based navigation in the oral cavity, close to the fiducial registration markers. Extraoral CBCT-guided navigation was performed for the retrieval of foreign bodies12 or orbital reconstruction.19 However, no evaluations of the accu-
Eggers et al. e45
Fig. 7. Distribution of the average target registration error in CBCT imaging. Because of the restricted scanning volume, the phantom skull is not imaged completely. The colored spheres indicate the locations of the target markers. The color indicates the average TRE for all registrations at the respective target marker.
Fig. 8. Distribution of the average target registration error (TRE) in CT imaging. The colored spheres indicate the locations of the target markers. The color indicates the average TRE. Small white spheres indicate target markers that were not evaluated because they were not visible in the corresponding CBCT image data set.
racy exist. Since the distance to the registration markers is larger, higher inaccuracies are to be expected.16 Hence, it was the objective of this study to determine the accuracy that can be achieved for CBCT-based navigation in midface and skull base and compared to the gold standard CT. Previous measurements indicated that the geometric accuracy of the Newtom 9000 DVT is slightly lower than that of CT, particularly in the marginal regions of the imaging volume. This is in accordance with pub-
e46
Eggers et al.
Fig. 9. Target registration errors at the various marker positions in CT and CBCT image data. Each dot indicates the TRE at one marker. The horizontal lines indicate the average TRE for CT and CBCT image data, respectively.
Fig. 10. Correlation analysis of registration accuracy. Each ⫹ indicates the TRE in CT (x-axis) and CBCT (y-axis) of a target marker position.
lished reports of higher inaccuracy of geometric measurements in marginal regions of CBCT image data.20 With the registration experiments we wanted to specify the resulting targeting accuracy when using CBCT image data. Since the imaging volume of the used device is limited, it is most likely that registration markers are located at the margin of the volume imaged. Hence the image data used for this study is a
OOOOE March 2009
less-than-optimum, though realistic configuration, for the position of the registration markers in image data for image guided maxillofacial surgery. The results of the registration experiments showed that there was no significant difference in the targeting accuracy when CT or CBCT image data were used. The average TRE values, which indicate the accuracy with which the surgeon is guided by the navigation system, were similar. Furthermore, when comparing the imaging modalities, there was a high correlation of TRE for every single target in CT and CBCT. This indicates that the surgeon can expect to have the same navigation accuracy based on CBCT image data that he or she would have had using CT instead. Hence it can be concluded, that the observed irregularities in the geometric accuracy of CBCT image data do not affect the registration accuracy. This can be explained by the fact that there are many more sources of error in imageguided surgery.21 Contributing to the overall error in navigation accuracy are the relocation error of the registration template, which is usually temporarily removed between imaging and surgery,22 inaccuracies of the tracking system,23 erroneous identification of the fiducial markers in image data and in patient space, and the geometric configuration of the registration markers and the position of a target relatively to the registration markers.24 From the technical point of view the use of CBCT image data for navigation is a good alternative to CT. Besides similar registration accuracy, there is a lower radiation dose6 and fewer dental metal artifacts.7 A general problem of phantom studies is the validity of the results on a real patient. The key issue in this study are the geometric distribution and identification of markers in a plastic phantom skull setting. Previous studies showed consistently good geometric accuracy of CBCT image data in plastic phantoms as well as on human.25,26 The slight decay of Newtom CBCT geometric accuracy to the margins of the volume of interest was found in plastic phantom studies as well as in human skull studies.15,20 Furthermore, we found a good representation of the registration markers in the phantom CT and CBCT image data, similar to clinical CT and CBCT image data (Fig. 6). Hence, we can assume that the use of a plastic phantom did not compromise the results. However, a problem that could not be addressed in this study was the problem of motion artifacts. There is always some motion of the patient during image data.27 This can reduce navigation accuracy.28 However, there are possible clinical circumstances that might restrict the use of CBCT. First, situations where high soft tissue contrast or the use of contrast enhancement agents is required. Here CT has better
OOOOE Volume 107, Number 3
diagnostic image quality. Another restriction might be the limited field of view of CBCT scanners in comparison to CT. The current development of CBCT imaging hardware shows 2 trends. First, by using larger detector units (e.g., 12 inches diameter instead of 9 inches in this study), imaging systems with larger or variable field of view become available.6 This mitigates the problem of restricted field of view for navigation purposes. Second, there is a change in detector technology. Flat-panel detectors (FPD) became more and more widespread as an alternative to the image intensifier (II) technology used in this study.29,30 Comparative studies in a variety of settings have shown that FPD devices can provide image data with less noise,31 fewer artifacts,32 lower radiation dose,33 and higher spatial resolution.34 However, FPD technology is still developing. Experience from cardiac imaging shows that FPD systems can be, but not necessarily are, better than II systems. Other components like x-ray tube, generator, and dose control contribute to image quality.35 There exists no definition of minimum navigation accuracy in literature. For paranasal sinus and anterior skull base surgery system accuracies of 1 to 2 mm were labeled “acceptable.”36 In a comparison of 5 navigation systems, accuracies of 1.76 to 3.22 mm were found to have “reached the high standard of precision necessary for interventions in clinical use.”37 An accuracy of up to 2 mm increased safety for patient and surgeon in lateral skull base surgery.38 The accuracy values found in this study show the equivalent accuracy of CBCT-based navigation as compared to the gold standard CT-based navigation. Whether the accuracies are sufficient or not depends on the individual surgical task. The results may at best support the surgeon in this decision by providing concrete numbers on the expected inaccuracy. CONCLUSION Registration of CBCT image data to the patient’s body for image-guided surgery is possible with similar geometric accuracy as that of CT. Hence, this imaging modality is appropriate as a basis for image-guided surgery of the face and anterior skull base. REFERENCES 1. Horsley VA, Clarke RH. The structure and functions of the cerebellum examined by a new method. Brain 1908;31:45-124. 2. Al-Rodhan NR, Kelly PJ. Pioneers of stereotactic neurosurgery. Stereotact Funct Neurosurg 1992;58:60-6. 3. Lunsford LD, Martinez AJ, Latchaw RE. Stereotaxic surgery with a magnetic resonance- and computerized tomography-compatible system. J Neurosurg 1986;64:872-8. 4. Eggers G, Mühling J, Marmulla R. Image-to-patient registration techniques in head surgery. Int J Oral Maxillofac Surg 2006;35: 1081-95.
Eggers et al. e47 5. Mischkowski RA, Zinser MJ, Ritter L, Neugebauer J, Keeve E, Zöller JE. Intraoperative navigation in the maxillofacial area based on 3D imaging obtained by a cone-beam device. Int J Oral Maxillofac Surg 2007;36:687-94. 6. Ludlow JB, Davies-Ludlow LE, Brooks SL, Howerton WB. Dosimetry of 3 CBCT devices for oral and maxillofacial radiology: CB Mercuray, NewTom 3G and i-CAT. Dentomaxillofac Radiol 2006;35:219-26. 7. Holberg C, Steinhäuser S, Geis P, Rudzki-Janson I. Cone-beam computed tomography in orthodontics: benefits and limitations. J Orofac Orthop 2005;66:434-44. 8. Wörtche R, Hassfeld S, Lux CJ, Müssig E, Hensley FW, Krempien R, et al. Clinical application of cone beam digital volume tomography in children with cleft lip and palate. Dentomaxillofac Radiol 2006;35:88-94. 9. Müssig E, Wörtche R, Lux CJ. Indications for digital volume tomography in orthodontics. J Orofac Orthop 2005;66:241-9. 10. Schulze D, Blessmann M, Pohlenz P, Wagner KW, Heiland M. Diagnostic criteria for the detection of mandibular osteomyelitis using cone-beam computed tomography. Dentomaxillofac Radiol 2006;35:232-5. 11. Van Assche N, van Steenberghe D, Guerrero ME, Hirsch E, Schutyser F, Quirynen M, et al. Accuracy of implant placement based on pre-surgical planning of three-dimensional cone-beam images: a pilot study. J Clin Periodontol 2007;34:816-21. 12. Pohlenz P, Blessmann M, Blake F, Heinrich S, Schmelzle R, Heiland M. Clinical indications and perspectives for intraoperative cone-beam computed tomography in oral and maxillofacial surgery. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2007;103:412-7. 13. Eggers G, Mühling J, Marmulla R. Template-based registration for image-guided maxillofacial surgery. J Oral Maxillofac Surg 2005;63:1330-6. 14. Stratemann SA, Huang JC, Maki K, Miller AJ, Hatcher DC. Comparison of cone beam computed tomography imaging with physical measures. Dentomaxillofac Radiol 2008;37:80-93. 15. Eggers G, Klein J, Welzel T, Mühling J. Geometric accuracy of digital volume tomography and conventional computed tomography. Br J Oral Maxillofac Surg 2008;46:639-44. 16. West JB, Fitzpatrick JM, Toms SA, Maurer CR Jr, Maciunas RJ. Fiducial point placement and the accuracy of point-based, rigid body registration. Neurosurgery 2001;48:810-6. 17. Schroeder W, Martin K, Lorensen W. The visualization toolkit, an object-oriented approach to 3D graphics. 3rd ed. Clifton Park, NY: Kitware Inc.; 2003. 18. Eggers G, Mühling J. Template-based registration for imageguided skull base surgery. Otolaryngol Head Neck Surg 2007;136:907-13. 19. Zizelmann C, Gellrich NC, Metzger MC, Schön R, Schmelzeisen R, Schramm A. Computer-assisted reconstruction of orbital floor based on cone beam tomography. Br J Oral Maxillofac Surg 2007;45:79-80. 20. Lascala CA, Panella J, Marques MM. Analysis of the accuracy of linear measurements obtained by cone beam computed tomography (CBCT-NewTom). Dentomaxillofac Radiol 2004;33: 291-4. 21. Ecke U, Maurer J, Boor S, Khan M, Mann WJ. Fehlerquellen der Navigation in der lateralen Schädelbasischirurgie. Darstellung von Einflussfaktoren in der Praxis. HNO 2003;51:386-93. 22. Fitzpatrick JM, Balachandran R, Labadie RF. Bite-block relocation error in image-guided otologic surgery. In: Barillot C, Haynor DR, Hellier P, editors. MICCAI 2004, Lect Notes Comput Sci 3217. Berlin Heidelberg: Springer; 2004. p. 518-25. 23. Morris A. Bone registration and tracking using an optical track-
e48
24.
25.
26.
27.
28.
29.
30.
31.
32.
OOOOE March 2009
Eggers et al.
ing system [master’s thesis]. Baltimore: Johns Hopkins University; 2001. Fitzpatrick JM, West JB, Maurer CR Jr. Predicting error in rigid-body point-based registration. IEEE Trans Med Imaging 1998;17:694-702. Ludlow JB, Laster WS, See M, Bailey LJ, Hershey HG. Accuracy of measurements of mandibular anatomy in cone beam computed tomography images. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2007;103:534-42. Stratemann SA, Huang JC, Maki K, Miller AJ, Hatcher DC. Comparison of cone beam computed tomography imaging with physical measures. Dentomaxillofac Radiol 2008;37:80-93. Bontempi M, Bettuzzi M, Casali F, Pasini A, Rossi A, Ariu M. Relevance of head motion in dental cone-beam CT scanner images depending on patient positioning. Int J CARS 2008; 3:249-55. Marmulla R, Mühling J. The influence of computed tomography motion artifacts on computer-assisted surgery. J Oral Maxillofac Surg 2006;64:466-70. Honey OB, Scarfe WC, Hilgers MJ, Klueber K, Silveira AM, Haskell BS, et al. Accuracy of cone-beam computed tomography imaging of the temporomandibular joint: comparisons with panoramic radiology and linear tomography. Am J Orthod Dentofacial Orthop 2007;132:429-38. Guerrero ME, Jacobs R, Loubele M, Schutyser F, Suetens P, van Steenberghe D. State-of-the-art on cone beam CT imaging for preoperative planning of implant placement. Clin Oral Investig 2006;10:1-7. Baba R, Ueda K, Okabe M. Using a flat-panel detector in high resolution cone beam CT for dental imaging. Dentomaxillofac Radiol 2004;33:285-90. Katsumata A, Hirukawa A, Okumura S, Naitoh M, Fujishita M,
33.
34.
35.
36.
37.
38.
Ariji E, et al. Effects of image artifacts on gray-value density in limited-volume cone-beam computerized tomography. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 2007;104:829-36. Hatakeyama Y, Kakeda S, Ohnari N, Moriya J, Oda N, Nishino K, et al. Reduction of radiation dose for cerebral angiography using flat panel detector of direct conversion type: a vascular phantom study. AJNR Am J Neuroradiol 2007;28:645-50. Baba R, Konno Y, Ueda K, Ikeda S. Comparison of flat-panel detector and image-intensifier detector for cone-beam CT. Comput Med Imaging Graph 2002;26:153-8. Davies AG, Cowen AR, Kengyelics SM, Moore J, Sivananthan MU. Do flat detector cardiac x-ray systems convey advantages over image-intensifier-based systems? Study comparing x-ray dose and image quality. Eur Radiol 2007;17:1787-94. Cartellieri M, Vorbeck F. Endoscopic sinus surgery using intraoperative computed tomography imaging for updating a threedimensional navigation system. Laryngoscope 2000;110:292-6. Cartellieri M, Kremser J, Vorbeck F. Comparison of different 3D navigation systems by a clinical “user.” Eur Arch Otorhinolaryngol 2001;258:38-41. Caversaccio M, Bächler R, Lädrach K, Schroth G, Nolte LP, Häusler R. The “Bernese” Frameless optical computer aided surgery system. Comp Aid Surg 1999;4:328-34.
Reprint requests: Georg Eggers, MD, DMD Department of Oral and Cranio-Maxillofacial Surgery Heidelberg University Hospital Im Neuenheimer Feld 400 69120 Heidelberg, Germany [email protected]