Intraoperative computed tomography and ... - Semantic Scholar

Dentomaxillofacial Radiology (2009) 38, 28–33 ’ 2009 The British Institute of Radiology http://dmfr.birjournals.org

RESEARCH

Intraoperative computed tomography and automated registration for image-guided cranial surgery G Eggers*,1, B Kress2, S Rohde2 and J Mu¨hling1 1

Department of Oral and Cranio-Maxillofacial Surgery, Heidelberg University Hospital, Germany; 2Department of Neuroradiology, Heidelberg University Hospital, Germany

Objectives: Two key problems for the use of navigation systems in image-guided surgery are accurate patient-to-image registration and the fact that with ongoing surgery the patient’s anatomy is altered while the image data remains unchanged. A system for intraoperative CT imaging and fully automated registration of this image addresses both problems. It had been evaluated successfully in phantom studies. In this clinical study, we assessed the impact of the system on intraoperative workflow and registration accuracy in everyday patient care. Methods: In ten patients who underwent image-guided surgery, CT image data were acquired intraoperatively and were automatically registered in the navigation system. Registration accuracy and surgical outcome were assessed clinically. In six of these patients, a maxillary splint with markers had been inserted to cross-check registration accuracy. The target registration error of these markers was measured. Results: In all cases, registration accuracy was clinically sufficient and the surgical task could be performed successfully. In those cases where a maxillary template with target markers was attached for additional control of the registration accuracy, the target registration error was always better than 2 mm. Automated registration reduced the intraoperative registration time considerably and partially compensated for the time needed to perform the image data acquisition. Conclusions: Intraoperative CT imaging and automated registration successfully address the two key problems of image-guided surgery. The method is robust and accurate and proved its usability in everyday patient care. Dentomaxillofacial Radiology (2009) 38, 28–33. doi: 10.1259/dmfr/26098099 Keywords: computed tomography, X-ray; neuronavigation

Introduction In standard image-guided surgery, the surgeon is informed in real-time about the position of an instrument relative to the patient’s body. The computer screen of the navigation system shows the tool-tip as image data for that patient. This requires two steps prior to the use of the navigation system. In the first step, acquisition of image data, e.g. CT, is necessary pre-operatively. The problem is that with the advancement of the surgical procedure, the image data no longer reflect the real anatomical situation.1 In the next step, the image data have to be aligned with the position of the patient in the operating room in a registration *Correspondence to: Georg Eggers, MD, DMD, Department of Oral and CranioMaxillofacial Surgery, Heidelberg University Hospital, Im Neuenheimer Feld 400, 69120 Heidelberg, Germany; E-mail: [email protected] Received 8 October 2007; revised 11 February 2008; accepted 15 February 2008

procedure. Registration is usually performed by manually identifying corresponding points on the patient’s body and as image data. This method, while established and reliable, is tedious, time-consuming and susceptible to errors by the surgeon resulting in inaccurate navigation.2 At our institution we have evaluated a system that was established to address both problems. It allows intraoperative CT image data acquisition to ensure congruency of surgical sites and image data. Furthermore, registration is performed automatically to reduce time and error. This paper reports our experience with the system’s registration accuracy against the background of registration accuracy requirements for image-guided surgery. Furthermore, we report the impact of this method on the integration of intraoperative imaging into surgical workflow.

Automated intraoperative CT registration G Eggers et al

Materials and methods All patients in this study were treated in our imageguided surgery operating room with an integrated CTimaging suite, an operating table for intraoperative imaging, a ceiling-mounted navigation system and connectivity equipment for these components. Operating table The operating table for intraoperative CT imaging (AWIGS; Maquet, Rastatt, Germany) is based on a metal structure. The patient lies on a radiolucent board, made of carbon fibre. For CT imaging, the patient can be slid from the table on this board. Imaging system The intraoperative CT imaging suite has a gantry (Emotion; Siemens, Forchheim, Germany) mounted on rails in the floor of the operating room. During surgery, the gantry is parked close to the wall of the operating room (Figure 1a). For imaging, the rails in the floor allow movement in a caudocranial direction. After the patient is slid from the operating table on the transfer board, the gantry is moved towards the patient to the most caudal position needed for imaging (Figure 1b). Thus it is ensured that no collision between the gantry and the structure of the operating table occurs. Scoutview and CT image data are then acquired in caudocranial direction. For automated registration, the housing of the gantry is equipped with reflective markers for infrared light. Hence its position can be measured for registration purposes. Navigation system Mounted to the ceiling of the operating room is the navigation system2 (VectorVision Sky; Brainlab, Heimstetten, Germany). A flat panel monitor shows image data to the surgeon and allows user interaction

a

29

via touch screen functionality. An infrared camera measures the positions of the patient and the surgeon’s instruments during navigation.3 For this purpose, passive tracking bodies with reflective spheres are attached to the patient as well as to the instrument to be tracked. Infrared light, emitted by the camera, is reflected and the position and orientation of the respective tracking body is measured. During surgery, the position of the instrument is displayed in the image data on the flat-panel monitor in real-time. Furthermore, this camera tracks the position of the CT gantry for registration purposes. In addition to the standard navigation software, the system is equipped with software to automatically receive and register intraoperatively acquired CT image data. Connectivity For the communication between the imaging system and the navigation system, a Digital Imaging and Communications in Medicine (DICOM) compatible network connection is implemented. The intraoperatively acquired CT image data can be transferred immediately to the navigation system, and are registered immediately and fully automatically. This process of transfer and registration takes about 1 min, depending on the number of image slices. Furthermore, the current position of the gantry in the z-axis (along the rails in the floor longitudinal to the patient) is measured and transferred to the navigation system Concept and workflow of automated registration In standard image-guided surgery, a registration procedure is necessary because the position of the patient in the surgical room and the position in the image data coordinate system are arbitrary. The basic idea of automated registration is to measure the positions of patient and gantry during image data acquisition. For these measurements the tracking

b

Figure 1 (a) Surgical position: A, the CT scanner is parked at the wall of the operating room; B, the operating table with C, the transfer board is at sufficient distance to provide enough room for the surgical team. (b) Imaging position: B, the operating table was moved towards A, the CT scanner; C, the radiolucent transfer board is slid out towards the gantry. The gantry is moved caudally for imaging Dentomaxillofacial Radiology


30

a

b

Figure 2 (a) Attachment of a dynamic reference frame prior to imaging. (b) A definite geometry of reflective markers serves as dynamic reference frame for the CT gantry

camera of the navigation system is used. From the position of the gantry, the position of the image plane is also known. Hence patient coordinates and image data coordinates are easily transformed into one coordinate system. Since the position of the patient is measured using the infrared tracking system of the navigation system, the dynamic reference frame is attached to the patient’s head prior to image data acquisition (Figure 2a). It is the same reference frame that is later used for navigation. For the measurement of the CT scanner position, a dynamic reference frame consisting of a definite geometric assembly of four reflective markers is attached to the cover of the gantry (Figure 2b). The geometry of the markers is calibrated to the spatial position of the CT image plane. Since this relation is invariant, the calibration had to be performed only once during installation of the system. With the known position of the image plane and the known position of the patient, all necessary information for the registration of the image data is present. Registration is performed immediately after transfer of the image data to the navigation system. Accordingly, the clinical workflow consists of the following steps:

N

N N N

N N N

N N

N N

a dynamic reference frame is attached to the patient’s head the operating table is moved automatically to a preset imaging position that aligns the transfer board with the opening of the gantry. The patient is slid out with the transfer board towards the gantry. To maintain sterility of the surgical field, a sterile drape is put on top of the patient while leaving the dynamic reference frame visible to the camera the instrument tables and the anaesthesiological equipment are moved away from the patient to avoid collisions with the gantry. Hence longer respirator tubes and intravenous lines are necessary the gantry is now moved towards the patient until the image plane is in the centre of the surgical region of

Dentomaxillofacial Radiology

N

interest. Since the highest navigation accuracy is desired here, this position is used as the basis for registration the infrared tracking camera is adjusted so that the dynamic reference frames of patient and gantry are in the field of view simultaneously (Figure 3). These positions are recorded by the system. From now on, the patient must not be moved any more! the gantry is now moved to the most caudal imaging position all personnel except the patient leave the room acquisition of a CT-based 3D image data set in the caudocranial direction. Patient and gantry do not have to be visible to the tracking camera any more. Motion of the gantry along the longitudinal (z) axis is measured by the CT system and is sent via a direct link to the navigation system. The image data is reconstructed and sent to the navigation system as DICOM data via network the personnel re-enter the operating room patient-to-image registration is performed automatically in the navigation system. The surgeon performs an accuracy check on anatomical landmarks using a tracked pointer if registration accuracy is questionable, a re-registration can be performed as long as the patient was not moved: therefore, the CT gantry and infrared camera are adjusted into the field of view of the tracking camera. The transformations then can be re-calculated for the already transferred image data after successful registration, the patient is slid back onto the operation table and the gantry is moved away. The operating table can be driven automatically to the position that was saved prior to imaging and surgery continues.

This system was used on ten patients for automated registration. After registration, a clinical accuracy check, based on anatomical landmarks and performed before surgery, was continued. Finally, we evaluated


31

the corresponding point in image data was measured as the target registration error (TRE) (Figure 4b).

Results

Figure 3 Preparation for image data acquisition: A, the infrared tracking camera was adjusted to record the positions of the tracking markers on B, the patient and C, the CT gantry. D, The navigation system is ready to receive the image data for automated registration

whether the surgical task could be completed successfully using the navigation information. In six patients, we performed an additional measurement of registration accuracy. We equipped the maxilla with a supplementary splint with titanium screws as target markers prior to CT image data acquisition (Figure 4a). These splints were made from a light curing resin (Triad Gel; Dentsply, York, PA) and were equipped with titanium screws (Stryker Leibinger Micro Implants, Freiburg, Germany) as target markers. After automated registration, the titanium screws were located on the patient using the tracked pointer of the navigation system. The distance of the pointer’s tip to

a

The automated registration of the intraoperative CT image data has been used in ten patients so far. The indications were either the removal of foreign bodies or the resection or biopsy of a space occupying lesion (Table 1). Image data were acquired as native CT images for the cases dealing with foreign bodies, and after administration of contrast agent in cases of tumour resection or biopsy. In all cases, the registration procedure was uneventful. After acquisition, the image data were sent to the navigation system via the network, and registration was performed fully automatically. Clinical accuracy was checked by touching distinct anatomical landmarks with the pointer while verifying the location on the image data set on the navigation system’s display. In all cases the accuracy appeared clinically to be in the order of magnitude of 1 mm to 2 mm. In six patients, the TRE of the attached titanium screw markers was measured using a maxillary splint with target markers. The average TRE for each of these six registrations was between 0.50 mm and 1.57 mm (Table 2). The overall average TRE of all markers in all registrations was 1.20¡0.48 mm (mean¡standard deviation) with a 95% confidence interval of 0.97–1.43 mm. Clinically, navigation was successful in all cases: the navigation system provided accurate and error-free information to locate and retrieve the foreign bodies

b

Figure 4 (a) Attachment of a maxillary splint with marker screws for accuracy measurements. (b) Measurement of target registration error after automated patient-to-image registration: the screenshot shows a triplanar view with axial (top right), coronal (bottom right) and sagittal (bottom left) projection. In each projection, the thin dotted crosshairs indicate the position of a target marker in image data (the centre of the head of a titanium screw). The conical tip (arrow) in the centre of the green crosshairs indicates the position of the same target marker in the patient, measured with the tracked pointer of the navigation system. The target registration error is the Euclidean distance between these two points (in this case 1.7 mm) Dentomaxillofacial Radiology


32 Table 1

Patients who underwent automated registration for intraoperative CT image data acquisition

No.

Gender

Age (years)

Imaging

Surgical task

Region

1 2 3 4 5 6 7 8 9 10

F F M F M M F M M M

47 43 31 20 80 71 33 40 63 68

Native Contrast Native Native Contrast Native Native Contrast Contrast Contrast

Foreign body removal Tumour resection Foreign body removal Foreign body removal Tumour resection Foreign body removal Foreign body removal Tumour biopsy Tumour biopsy Tumour biopsy

Midface Cheek Cheek Midface Face Midface Midface Orbit Tongue base Parapharyngeal

Imaging: Contrast, the use of an intravenous contrast agent for imaging (Ultravist 300; Bayer Vital GmbH, Leverkusen, Germany); Native, native CT image data acquisition

through minimal incisions. In all cases of tumour therapy the guidance of the system allowed for successful biopsy or resection of the space occupying lesions. The impact of the system on intraoperative workflow was twofold: on the one hand, surgical workflow was impaired by the necessary preparations for image data acquisition. The necessary steps of the workflow would interrupt surgery for approximately 30 min. However, there was a learning curve in the handling of the set-up and the patient, with an ongoing tendency to shorter interruptions. On the other hand, automated registration made any kind of registration procedure unnecessary as, for example, the identification of anatomical or artificial points for pair-point registration or surface data acquisition for a surface matching procedure.

Discussion The foundations for intraoperative imaging and the use of this data for image-guided surgery were laid in neurosurgery.4 The use of intraoperative stereotactic X-ray equipment and the use of this data for frame-based stereotactic interventions are the first applications in this direction. Also, intraoperative volume imaging and its use for imageguided surgery were first introduced in neurosurgery with dedicated MRI.5,6 Intraoperative CT imaging is comparably young and there are only a few reports about its use in maxillofacial surgery (e.g. see references7,8). Table 2 Accuracy of the automated registration, as measured on target markers on a maxillary splint Target registration error (mm) Number of target Patient no. markers measured

Average

Standard deviation

1 2 3 4 5 6

1.30 1.57 1.40 1.27 0.50 1.15

0.20 0.50 0.17 0.31 0.44 0.07

3 3 3 3 3 2*

*Only two target markers were visible in the image data set. The third marker was outside the selected field of view Dentomaxillofacial Radiology

An inherent disadvantage of intraoperative CT imaging is the interruption of the surgical procedure.9 While the workflow for the image data acquisition is quite straightforward, there is still an interruption time of about 30 min for the imaging procedure. These 30 min are precious operating room time. Hence intraoperative imaging is not a routine method.10 Preoperative imaging, if sufficient, is preferable for economical reasons. However, the key advantage of intraoperative imaging is the fact that the surgeon can have image information of the current intraoperative situation and is not forced to rely on pre-operative image data. This becomes particularly important when the image data are a key element of orientation as in image-guided surgery. However, the intraoperatively acquired image data are not available for navigation before they were registered to the patient. While existing methods like pair-point registration are accurate as long as basic principles and constraints are observed, they require additional interaction by the surgeon. Corresponding registration points have to be identified in image data and on the patient’s body, and the resulting navigation accuracy is distributed inhomogeneously over the volume of the head.11 In contrast, in a previous phantom study the distribution of the TRE using this automated registration system was more homogeneous and the overall accuracy was very good with average TRE values below 1.5 mm. However, in that study the target markers were distributed on all regions of a plastic phantom skull and not only in the oral region. Possible clinical sources of error like inadvertent motion of the patient during imaging, or dislocation of the marker template could not occur.12 The other advantage is that the automated registration saves time because the necessary calculations do not cause a noticeable delay. After image data transfer, image data are registered immediately. In those six patients where we performed a measurement of registration accuracy on maxillary mounted target markers, the overall accuracy was in the same order of magnitude, below 1.5 mm. There are, however, no universally accepted accuracy limits for image-guided surgery because the necessary accuracy will always depend on the surgical task. There are publications on image-guided surgery where an


accuracy of 1.5 mm was deemed helpful for surgical tasks like skull base surgery.10,13,14 We used the imaging and registration system for two kinds of applications: the retrieval of foreign bodies and the biopsy or resection of space occupying lesions. In both surgical tasks, the role of the system is to guide the surgeon to a designated area of interest. This worked out in all cases. Biopsies were performed successfully and all foreign bodies could be retrieved. For the tumour resection, the resection margins could be planned based on the intraoperative imaging and executed using the guidance of the navigation system. The fact that the image data were acquired intraoperatively is particularly useful in cases where the location of the region of interest would change with the position of the head or the mandible. This becomes relevant when the region of interest is located in soft tissues that are shifted with changes in neck or mandible position, because the positions of the patient’s mandible and neck are not the same in pre-operative CT imaging as in the operating room.15 Registration of preoperative image data would bear the danger that the

33

region of interest (e.g. the tumour), indicated by the navigation system on pre-operative image data set, is somewhere else due to the soft tissue shift, e.g. in cases with foreign bodies in the cheek or biopsies in the base of the tongue or the pharynx. The same is true for mobile foreign bodies. Furthermore, intraoperative changes caused by the surgeon are reflected, e.g. in the intraoperative imaging for resection control in tumour surgery.16 In conclusion, intraoperative CT imaging is a valuable tool, particularly in image-guided surgery. The trade-off is the interruption of the surgical procedure for the image data acquisition. The fully automated registration is a highly accurate registration method that enhances the clinical workflow: as soon as the image data is transferred, image-guided surgery can continue without further delay. Thus the integration of intraoperative imaging is improved. Based on the results of this study, the system for automated registration is now in routine use in patient care at our institution. It is used in cases of image-guided surgery on intraoperatively acquired CT image data.

References 1. Heiland M, Habermann CR, Schmelzle R. Indications and limitations of intraoperative navigation in maxillofacial surgery. J Oral Maxillofac Surg 2004; 62: 1059–1063. 2. Eggers G, Mu¨hling J, Marmulla R. Image-to-patient registration techniques in head surgery. Int J Oral Maxillofac Surg 2006; 35: 1081–1095. 3. Wiles AD, Thompson DG, Frantz DD. Accuracy assessment and interpretation for optical tracking systems. In: Galloway Jr RL (ed). Medical imaging 2004: visualization, image-guided procedures, and display. proceedings of SPIE Vol. 5367. Bellingham, WA: SPIE, 2004, pp 421–432. 4. Horsley V, Clarke RH. The structure and functions of the cerebellum examined by a new method. Brain 1908; 31: 45–124. 5. Black PM, Moriarty T, Alexander E 3rd, Stieg P, Woodard EJ, Gleason PL, et al. Development and implementation of intraoperative magnetic resonance imaging and its neurosurgical applications. Neurosurgery 1997; 41: 831–842. 6. Wirtz CR, Bonsanto MM, Knauth M, Tronnier VM, Albert FK, Staubert A, et al. Intraoperative magnetic resonance imaging to update interactive navigation in neurosurgery: method and preliminary experience. Comput Aided Surg 1997; 2: 172–179. 7. Hoffmann J, Dammann F, Reinert S. Initial experience with intraoperative computed tomography in maxillofacial surgery. Biomed Tech (Berl) 2002; 47S1: 470–473. 8. Ho¨lzle F, Klein M, Schwerdtner O, Lu¨th T, Albrecht J, Hosten N, et al. Intraoperative computed tomography with the mobile CT Tomoscan M during surgical treatment of orbital fractures. Int J Oral Maxillofac Surg 2001; 30: 26–31.

9. Stieve M, Schwab B, Haupt C, Bisdas S, Heermann R, Lenarz T. Intraoperative computed tomography in otorhinolaryngology. Acta Otolaryngol 2006; 126: 82–87. 10. Cartellieri M, Vorbeck F. Endoscopic sinus surgery using intraoperative computed tomography imaging for updating a three-dimensional navigation system. Laryngoscope 2000; 110: 292–296. 11. 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: 810–816. 12. Eggers G, Kress B, Mu¨hling J. Automated registration of intraoperative CT image data for navigated skull base surgery. Minim Invasive Neurosurg 2008; 51: 15–20. 13. Caversaccio M, Ba¨chler R, La¨drach K, Schroth G, Nolte LP, Ha¨usler R. Frameless computer-aided surgery system for revision endoscopic sinus surgery. Otolaryngol Head Neck Surg 2000; 122: 808–813. 14. Cartellieri M, Kremser J, Vorbeck F. Comparison of different 3D navigation systems by a clinical ‘‘user’’. Eur Arch Otorhinolaryngol 2001; 258: 38–41. 15. Hoffmann J, Troitzsch D, Westendorff C, Weinhold O, Reinert S. Temporary intermaxillary fixation using individualized acrylic splints permits image-data-based surgery of the lower jaw and oropharynx. Laryngoscope 2004; 114: 1506–1509. 16. Grunert P, Mu¨ller-Forell W, Darabi K, Reisch R, Busert C, Hopf N, et al. Basic principles and clinical applications of neuronavigation and intraoperative computed tomography. Comput Aided Surg 1998; 3: 166–173.

Dentomaxillofacial Radiology