Neurosurgery 68[ONS Suppl 1]:ons95 ons102, 2011

Transcription

1 STEREOTACTIC & FUNCTIONAL Operative Technique Target and Trajectory Clinical Application Accuracy in Neuronavigation Reuben R. Shamir, MSc* Leo Joskowicz, PhD* Sergey Spektor, MD Yigal Shoshan, MD *School of Engineering and Computer Science, The Hebrew University, Jerusalem, Israel; Department of Neurosurgery, The Hebrew University Hadassah Medical Center, Jerusalem, Israel Correspondence: Yigal Shoshan, MD, Department of Neurosurgery, Hadassah Hebrew University Medical Center, 91120, Jerusalem, Israel. Received, December 23, Accepted, August 12, Copyright ª 2011 by the Congress of Neurological Surgeons BACKGROUND: Catheter, needle, and electrode misplacement in navigated neurosurgery can result in ineffective treatment and severe complications. OBJECTIVE: To assess the Ommaya ventricular catheter localization accuracy both along the planned trajectory and at the target. METHODS: We measured the localization error along the ventricular catheter and on its tip for 15 consecutive patients who underwent insertion of the Ommaya catheter surgery with a commercial neuronavigation system. The preoperative computed tomography/magnetic resonance images and the planned trajectory were aligned with the postoperative computed tomography images showing the Ommaya catheter. The localization errors along the trajectory and at the target were then computed by comparing the preoperative planned trajectory with the actual postoperative catheter position. The measured localization errors were also compared with the error reported by the navigation system. RESULTS: The mean localization errors at the target and entry point locations were and mm, respectively. The mean shift and angle between planned and actual trajectories were mm and , respectively. The mean difference between the localization error at the target and entry point was mm. The mean difference between the target localization error and the reported navigation system error was mm. CONCLUSION: The catheter localization errors have significant variations at the target and along the insertion trajectory. Trajectory errors may differ significantly from the errors at the target. Moreover, the single registration error number reported by the navigation system does not appropriately reflect the trajectory and target errors and thus should be used with caution to assess the procedure risk. KEY WORDS: Computer assisted surgery, medical device safety, neuronavigation, Ommaya reservoirs Neurosurgery 68[ONS Suppl 1]:ons95 ons102, 2011 DOI: /NEU.0b013e d9 Precise targeting of tumors, lesions, and anatomic structures with a probe, needle, catheter, or electrode inside the brain based on neuronavigation with preoperative computed tomography/magnetic resonance imaging (CT/MRI) is the standard approach in many keyhole neurosurgical procedures. Imageguided neuronavigation systems enable the execution of minimally invasive procedures by showing the surgeon the actual location of the ABBREVIATIONS: ARE, angular registration error; ERE, entry registration error; SRE, shift registration error; TRE, target registration error handheld tool on the preoperative image augmented with the planned trajectory and target. A key step in neuronavigation is the accurate intraoperative alignment, commonly called registration, between preoperative CT/MRI images and the intraoperative physical head. A common approach for patient-to-image registration is point-based registration, in which points are individually selected on each data set and paired to each other on the basis of their correspondence. Points can be marked with bone-mounted screws/spheres, adhesive skin markers, and anatomic landmarks. It has been shown that bone-mounted screws result in better accuracy than adhesive skin markers and anatomic NEUROSURGERY VOLUME 68 OPERATIVE NEUROSURGERY 1 MARCH 2011 ons95

2 SHAMIR ET AL markers. 1,2 However, they are invasive, and with regard to diagnostic yield and complication rate, the frameless stereotactic biopsy procedure was found to be comparable to or better than the frame-based method. 3 Some studies have shown that adhesive skin markers are associated with better accuracy than anatomic landmarks. 4,5 However, we have shown that this is not always the case because the accuracy depends on the selected anatomic landmarks and on the number of markers and their locations. 6 The tragus, for example, is associated with a lower localization error than the adhesive marker. Moreover, we observed situations in which skin markers were dropped and reattached between preoperative scanning and the operation and situations in which the adhesive marker drifted during intraoperative registration because of skin stretching. In addition, adhesive markers require shaving larger areas and increase discomfort to the patient. Inaccuracies in target and trajectory localization when using image-guided neuronavigation systems may result in serious complications such as hemorrhage, neurological deficit, catheter malfunction, and nondiagnostic tissue samples. 3,7-9 Because of its clinical importance, the localization accuracy of image-guided navigation has been studied extensively. 1,2,4,6,10-30 The most clinically relevant measure is the discrepancy between the planned target location as defined on the preoperative CT/MRI and the actual tool tip location when it reaches the target with the navigation guidance. This discrepancy is called the target registration error (TRE). 31 In vivo studies usually report the fiducial registration error or the TRE on markers attached to the patient skull or estimate the intracranial accuracy with a theoretical formula or laboratory experiments. 4,6,15,17,24,25 Clinical studies, including our own, show that these error measures do not necessary reflect the actual intracranial accuracy. Only a few studies measured the actual intracranial TRE of the navigation systems by fusing the preoperative image and planned target with the postoperative image. 15,16,18,19,23 No study has been conducted to assess the localization error along the insertion trajectory of the catheter, needle, or probe and the related risk to nearby structures such as blood vessels. The aim of this study is to retrospectively quantify the localization error both at the target and along the entire insertion trajectory, in addition to the standard fiducial registration error/tre. The localization accuracy along a trajectory may have a high variability and may be significantly different from the TRE because the accuracy varies with the location of interest. 6,11,20,25,31,32 Specifically, we measure the localization errors along the trajectories and targets for patients who underwent insertion of the Ommaya catheter guided by a commercial neuronavigation system. Ommaya catheter insertion surgery may be indicated for onsite repeated drug delivery to the ventricular system, for repeated cerebrospinal fluid (CSF) sampling, and for cystic lesions evacuation The implanted catheter is clearly visualized on postoperative CTs, so its skull entry point, the catheter tip final location, and the entire trajectory can be measured and compared with the preoperative surgical planning (Figure 1A). CLINICAL MATERIAL AND METHODS We retrospectively measured the localization error along the catheter trajectory and on its tip for 15 consecutive patients who underwent FIGURE 1. Illustration of how the planned and the actual target and trajectory locations are compared with the Medtronic Cranial Application, A, alignment and fusion of the planned trajectory with the axial postoperative image showing the displacement between the actual catheter and its intended location and position. B, probe eye s view along the planned trajectory enables the computation of the angular and shift registration errors. ons96 VOLUME 68 OPERATIVE NEUROSURGERY 1 MARCH

3 APPLICATION ACCURACY IN NEURONAVIGATION keyhole minimally invasive Ommaya catheter placement surgery by 3 different neurosurgeons. Hardware and Instrumentation The optical digitizer-based neuronavigation system used in our procedures has been previously described (StealthStation/Treon, Medtronic, Boulder, Colorado). 1,23 The catheter guide, a hollow 120-cm-long tube with 4 LEDs mounted on a butterfly-shaped shield, is firmly attached to the surgical arm. We used a specially designed, adjustable, 3-piece surgical arm (Medtronic) that is attached to the Mayfield head holder and rigidly fixed during the procedure by way of single prominent thumbscrew, creating a stable platform that provides precise trajectory refinement along isolated axes. Surface facial landmarks and fiducial markers are the basis for point-to-point registration between the patient s head and the image volume. Nine surface facial anatomic landmarks and 1 fiducial applied above the bregma were used for registration. A typical setup incorporated the left/right tragus, left/right lateral and medial cantus, left/right helical cruz, and the nose bridge. Imaging Data Acquisition and Processing Preoperative, 3-dimensional (3D), contrast T1-weighted MRIs (GE Medical Systems, Milwaukee, Wisconsin) with 1-mm slide width containing the areas of interest and the set of anatomic and fiducial markers were obtained the night before surgery for 5 patients (33%). Ten patients (66%) completed preoperative 1.3-mm-slide-width contrast CT images for interpolated 3D volume visualization. Image resolutions are and mm 3, respectively. Imaging studies were transferred via the hospital network to the neuronavigation system and presented in triplanar anatomic and/or navigational views, as well as 3D surface reconstruction and oblique views. Surgical Planning The anatomic and fiducial markers were identified in the images, and their locations were recorded. The catheter tip target is selected just above the floor of the right frontal horn 5 mm anterior to the foramen of Monro to avoid adhesion of the choroid plexus with the catheter holes and to allow optimal location of all the catheter holes within the ventricular system. The preferred cortical entry point is chosen usually at the middle or superior frontal gyrus anterior to the coronal suture remote from CT/MRI-visible major cortical veins or the sagittal sinus. By using the multiplanar and oblique images and target and trajectory guidance features displayed on the computer screen, we can simulate each selected entry and trajectory to obtain the optimal surgical plan. On completion of the surgical planning process, the selected target and entry points are recorded and stored. Operative Technique The patient is placed in a supine position with the head in neutral position with slight flexion after full general anesthesia is administered. The patient s head is firmly fixed in a Mayfield head-holder attached to the surgical table. The reference arc is then clamped to the Mayfield head holder. Patient-to-image registration is performed with the neuronavigation wand and the 10 predefined surface facial landmarks and/or adhesive marker. After the registration is complete, we perform a short neuronavigation session for registration accuracy verification and locate the entry point previously selected in the surgical planning phase. The entry point is marked, minimally shaved, and prepared with antiseptic solution, and the patient s head is covered with a sterile drape. The multiarticulated arm is now mounted on the Mayfield head holder, and a sterile reference arc is placed. To provide stabilization, the catheter guide tip is positioned at the entry point flush to the skin surface and is activated and maneuvered until 3 colored circles representing the true target and the projection of the actual trajectory are displayed concentrically, indicating that the guide is located on the selected entry point and pointed toward the target in the proper trajectory. A 3-cm linear incision centered at the entry point is made, and the multiarticulated arm, with the attached catheter guide, is rigidly fixed against the skull with a single screw (Figure 2A). Position and trajectory alignment are verified again, and the exact catheter length from the tabula interna to the target is recorded and marked with a black sterile marker on a standard 2-mm ventricular catheter (Figure 2B). A handheld twist drill hole (2.7-mm drill diameter) is made, penetrating the skull inner table by a few millimeters to incise the dura. A 2.7-mm reducing tube is secured into the catheter guide. Then, the ventricular catheter is inserted to the FIGURE 2. Intraoperative execution. A, the catheter guide attached to the surgical arm is rigidly fixed against the skull, and the ventricular catheter is passed through a 2.7-mm reducing tube (not shown) to the target. B, after removal of the catheter guide, the specific catheter length is marked on a standard 2-mm ventricular catheter (black arrowhead) and passed through 2.7-mm twist drill. NEUROSURGERY VOLUME 68 OPERATIVE NEUROSURGERY 1 MARCH 2011 ons97

4 SHAMIR ET AL target with a stylet (Figure 2A). The stylet is then pulled out and CSF flow is verified. At this point, the catheter is secured manually against the skull with a coated bayonet while the reducing tube and the catheter guide are carefully removed (Figure 2B). The Ommaya reservoir is then attached to the catheter end and placed under the skin medial to the incision. The catheter black length mark location is verified exactly at the entrance to the skull (Figure 2B), and the reservoir is anchored with 3/0 silk suture to the periost. CSF sampling is obtained through the reservoir for routine analysis, and the skin is closed in normal fashion. A volume noncontrast CT scan with a resolution of mm 3 was performed later the same day for all patients to screen for catheter positioning before injection of chemotherapy and for silent hemorrhage. The Ommaya catheter is clearly visible on the postoperative image (Figure 1A). Patients with no clinical or radiological complications are discharged the next morning. Measurements To quantify the planned vs resulting trajectory accuracy, we performed an automatic rigid registration between the preoperative CT/ MRI image and the postoperative CT image of each patient with the Cranial Application (Medtronic) image-fusion module. The accuracy of the image fusion used in this study may have a direct effect on the localization error. Mascott s 16 study reports that the accuracy of the automatic registration method in the Medtronic station is practically identical to that of a manual registration method with skull implanted markers, which is, 1 mm on average. For each of the 15 cases, a senior neurosurgeon validated the image fusion visually on predefined anatomies and anatomic landmarks. In all cases, the registration was deemed visually accurate, with no deviations observed between the fused preoperative and postoperative images. To quantify the target and trajectory registration errors, we define 4 clinically relevant error measures: TRE, the distance between the planned and actual catheter tip location; entry registration error (ERE), the distance between the planned and actual catheter skull entry point; angular registration error (ARE), the angle between the planned and actual surgical tool trajectory; and shift registration error (SRE), the distance between the closest points on the planned and actual surgical tool trajectories. The SRE is defined because 2 trajectories in the 3D space do not necessarily intersect each other. When the actual trajectory intersects the planned one, the SRE is zero but the other errors are not. In addition, we recorded the reported navigation registration error from the navigation system when available. The centerline of the implanted catheter was accurately identified (SD, 0.3 mm) on the postoperative CT images by repetitive selections of the entry, target, and tens of points along the trajectory. It was hard to define the actual catheter insertion point because the catheter grey level in the image was the same as that of the skull. Detection of the actual entry point required further adjustments of the color parameters (eg, window, level, and color map) and careful visual inspection in the vicinity of the entry-point zones. Despite our efforts, in 3 of the 15 cases, the variability in the repetitive selections of the entry point was large (SD. 0.5 mm), so its location could not be identified with sufficient accuracy and was discarded. The spatial errors were then computed from the 2-dimensional measurements obtained on orthogonal planes as we describe in the following section (Figure 1). All 4 measures, ERE, SRE, ARE, and TRE, are needed to describe the discrepancy between two 3D spatial trajectories. We investigate the contribution of each component to get a broader picture regarding the actual application accuracy. Target and Entry Registration Errors Given the distance between the planned point of interest (eg, target or entry point) and its actual location on 2-dimensional perpendicular views of the 3D image, the spatial distance between the 2 points is derived by the following formula (Figure 3): 23spatial dist 2 ðp; q Þ ¼ plane xy dist 2 ðp; qþ 1 plane xz dist 2 ðp; qþ þ plane yz dist 2 ðp; qþ where p and q are the points of interest, spatial_dist is the desired spatial distance between the points, and plane_xy_dist, plane_xz_dist, and plane_yz_dist are the measured distances on the 2-dimensional orthogonal views of the spatial preoperative and postoperative images. Angular and Shift Registration Errors To measure angular and shift registration errors, we selected the probe s eye view mode in the Cranial Application. The probe s eye view is perpendicular to the planned trajectory (Figure 1B) and can be scrolled along the planned trajectory. Because the preoperative image (and hence the planned trajectory) is aligned with the postoperative image, we can directly measure the signed distances from the planned trajectory to the actual catheter placement at each point of the trajectory (when the actual catheter placement intersects the planned trajectory, we multiply by 1 the distances after the intersection). Then, a 2-dimensional line is best fitted in a least-squares manner on the measured data, and the angle between planned and actual trajectories is computed (Figure 4). The SRE is the distance between the closest points of the planned and actual trajectories. FIGURE 3. The spatial distance between the planned and actual entry and target points was computed from distances between the points measured on 3 orthogonal 2-dimensional views as follows. Let p = (x p,y p,z p ) be the planned location (black crossed circle) and q = (x q,y q,z q ) be the actual location (grey crossed circle). The distance between p and q is spatial_dist 2 (p,q) = (x p -x q ) 2 +(y p 2 y q ) 2 + (z p 2 z q ) 2. The distances on the planes xy, xz, and yz are as follows: plane_xy_dist 2 (p, q) = (x p 2 x q ) 2 +(y p 2 y q ) 2,plane_xz_dist 2 (p, q) = (x p 2 x q ) 2 + (z p 2 z q ) 2, and plane_yz_dist 2 (p, q) = (y p 2 y q ) 2 +(z p -z q ) 2, respectively. Therefore, the sum of the 3 squared 2-dimensional distances is twice the squared spatial distance (Eq. 1). ons98 VOLUME 68 OPERATIVE NEUROSURGERY 1 MARCH

5 APPLICATION ACCURACY IN NEURONAVIGATION FIGURE 4. The actual trajectory was computed with best-fit line of the measured signed distances ( x ) between the planned and actual trajectories. The angle between the planned and the actual trajectories is computed from the trajectories. RESULTS The Table summarizes the results. The mean TRE, ERE, ARE, and SRE were mm (maximum, 14.9 mm), mm (maximum, 11.5 mm), (maximum, 16.0 ), and mm (maximum, 7.2 mm), respectively. The mean navigation registration error was mm (maximum, 1.9 mm). The mean absolute difference between the TRE and ERE was mm (maximum, 10.9 mm). The mean absolute difference between the actual TRE and the navigation registration error was mm (maximum, 13.2 mm). TABLE. Summary of the Measured Target and Trajectory Registration Errors for 15 Patients Who Underwent Ommaya Catheter Placement a Patient TRE, mm ERE, mm ARE, degrees SRE, mm NRR, mm Average (SD) 5.9 (4.3) 3.3 (1.9) 3.9 (4.7) 1.6 (1.9) 1.4 (0.4) a ARE, angular registration error; ERE, entry registration error; NRR, navigation registration error; SRE, shift registration error; TRE, target registration error. DISCUSSION Image-guided surgery is a mature technology with clinically demonstrated benefits However, its full range of capabilities and limitations and its application accuracy are still under investigation. 4,6,40,41 The present study aims to measure the registration and application error of the entire trajectory from the entry point down to the target by comparing preoperative planning for Ommaya catheter insertion surgery with the actual catheter location on postoperative imaging. We found that the true average TRE was mm. Such error may result in misplacement of the catheter into the third ventricular or in the adjacent brain parenchyma. Dislocation of the ventricular catheter may also be associated with catheter malfunction. 42 Our results include the clinical application accuracy of the commercial navigation system. Extrapolating our results to the case of frameless stereotactic biopsy, this error may lead to nondiagnostic samples, especially in small-volume lesions, or to significant hemorrhage when the lesion is in close proximity to blood vessels. 7 Our observations also show that the localization error changes significantly along the trajectory. The TRE for patient 1 was 3.7 mm, which was deemed clinically acceptable. However, 52 mm before the target, at the cortical penetration point of the catheter, a catheter misplacement of 11.5 mm was measured, which is 3 times as much as the TRE error and may have clinical significance. The cortical surface is rich with blood vessels, so a deviation of such magnitude from the planned cortical penetration point might increase the risk of vascular injury. This shows that although the TRE is clinically acceptable, the localization error along the trajectory may not be adequate. In addition, low ARE and SRE do not necessarily imply low TRE. Patient 4 has ARE and SRE values that are much lower than those of patient 1, but the observed TRE of patient 4 was almost double that of patient 1. Therefore, the above parameters are not directly correlated to each other. It is important to consider the entire NEUROSURGERY VOLUME 68 OPERATIVE NEUROSURGERY 1 MARCH 2011 ons99

6 SHAMIR ET AL trajectory because critical structures such as blood vessels and eloquent areas can reside far from the target but close to the trajectory and the entry point. Another important observation is that the registration error reported by the navigation system may significantly differ from the actual trajectory and TREs. In patient 13, the navigation system reported a registration error of 1.9 mm, whereas the actual TRE was 13.4 mm. This finding is in line with the results of previous studies. 6,15,25 One reason for this underestimation is the existence of other sources of error that are not accounted for in patient-to-image registration. Additional sources of error in the localization of the catheter include the preoperative image quality and resolution, the accuracy of the tracking system, the tracked tool calibration error, and the catheter manipulation during the procedure. We observe from the fused images that the catheter is usually deeper than the planned target and believe that it was caused mainly during the fixation of the catheter and the attachment of the reservoir to it and the periost. No brain shift was observed in any of the cases. Moreover, the intracranial catheter remained straight (no bending) in the postoperative images of all patients, as can be seen in Figure 1 and as confirmed by our measurements (Figure 4). This result is consistent with the fact that the catheter is guided with a stylet that prevents its bending during insertion. After the stylet is removed, about 4 cm of the catheter remains within the supporting frontal lobe parenchyma; only the distal 2 to 2.5 cm of the catheter is located in the ventricle itself; and the catheter tip is usually positioned against the floor of the frontal horn. Therefore, in this situation, the catheter will largely remain straight as demonstrated on postoperative CT. In addition, recent theoretical studies show that even when considering only the internal sources of error in patient-to-image registration, common measures can underestimate the TRE. 40,41 Therefore, large discrepancies between the expected and actual TREs occur even without external sources of error that are not involved in the patient-to-image registration. The large measured TRE values suggest that the present method may not be safely used in applications that require a very precise localization. Moreover, it is clear that a single registration error value such as that reported by the navigation system cannot describe the large variability of errors along the trajectory. Therefore, other methods for the estimation and visualization of the system accuracy are needed. Because facial anatomic landmarks are sometime associated with less accurate registration than multiple adhesive skin markers, 5 it is possible that other registration setups will result in smaller error values that may be closer to the error reported by the navigation system. Yet, because angular error is also important, the single number reported by the commercial navigation system does not reliably represent the actual error in surgical tool insertion that varies along the trajectory. Therefore, the expected error should be with respect to a specific target location. We identify several ways for improving the accuracy and safety in neuronavigation systems. To improve accuracy, skull- or skinmounted markers can be used. 4 However, markers do not always lead to a clinically significant accuracy improvement. 11 Moreover, adhesive markers are sometimes associated with patient discomfort and may fall or shift because of mishandling. Surfacebased registration is associated with registration errors that are similar to those of anatomic landmarks. 4 Another option is to compute the optimal marker locations and the best anatomic landmarks subset. 10,43 These methods show the neurosurgeon what anatomic landmarks to choose and where to place the markers on the basis of the diagnostic images and with respect to the target location. Preliminary experiments show a significant improvement in the accuracy of tool localization. Moreover, new point-based registration methods 44,45 that incorporate a more realistic localization error model have the potential to better estimate the patient-to-image transformation. As for improving the evaluation of the expected accuracy, recent TRE estimation methods incorporate a more realistic localization error model and have the potential to better estimate the actual TRE. 46,47 Yet, none of the recent or previous methods were validated in a realistic clinical neuronavigation setup. We are currently developing an uncertainty geometric model that is based on the measured errors and that can be used for the evaluation of possible surgery outcomes. Methods for the computation of trajectories that are in a safe distance from critical structures reduce the risks associated with tool misplacement in neuronavigation. 48 The methods automatically compute the risks along tens of thousands of candidate trajectories and present them to the neurosurgeon for further refinement. Methods to improve the accuracy and safety of the trajectory and target localization should be further developed and examined under a clinical setup. CONCLUSION The catheter localization errors in standard neuronavigation have significant variations at the target and along the insertion trajectory, which may differ significantly from the error at the target. Furthermore, the single registration error number reported by the commercial navigation system does not appropriately reflect the trajectory and target error and thus should be used with caution to assess the risk of the procedure. The neurosurgeon should consider the optional significant deviation from the planned trajectory and the consequent outcome complications when selecting the entry and target points. Disclosure This work was supported by FP7 ERC ROBOCAST grant The authors have any no personal financial or institutional interest in any of the drugs, materials, or devices described in this article. REFERENCES 1. Bjartmarz H, Rehncrona S. Comparison of accuracy and precision between framebased and frameless stereotactic navigation for deep brain stimulation electrode implantation. Stereotact Funct Neurosurg. 2007;85(5): Henderson JM, Holloway KL, Gaede SE, Rosenow JM. The application accuracy of a skull-mounted trajectory guide system for image-guided functional neurosurgery. Comput Aided Surg. 2004;9(4): ons100 VOLUME 68 OPERATIVE NEUROSURGERY 1 MARCH

7 APPLICATION ACCURACY IN NEURONAVIGATION 3. Woodworth GF, McGirt MJ, Samdani A, Garonzik I, Olivi A, Weingart JD. Frameless image-guided stereotactic brain biopsy procedure: diagnostic yield, surgical morbidity, and comparison with the frame-based technique. J Neurosurg. 2006;104(2): Woerdeman PA, Willems PW, Noordmans HJ, Tulleken CA, van der Sprenkel JW. Application accuracy in frameless image-guided neurosurgery: a comparison study of three patient-to-image registration methods. J Neurosurg. 2007;106(6): Villalobos H, Germano IM. Clinical evaluation of multimodality registration in frameless stereotaxy. Comput Aided Surg. 1999;4(1): Shamir RR, Joskowicz L, Spektor S, Shoshan Y. Localization and registration accuracy in image guided neurosurgery: a clinical study. Int J Comput Assist Radiol Surg. 2009;4(1): Dammers R, Haitsma IK, Schouten JW, Kros JM, Avezaat CJ, Vincent AJ. Safety and efficacy of frameless and frame-based intracranial biopsy techniques. Acta Neurochir (Wien). 2008;150(1): Joskowicz L, Shamir R, Freiman M, et al. Image-guided system with miniature robot for precise positioning and targeting in keyhole neurosurgery. Comput Aided Surg. 2006;11(4): Favre J, Taha JM, Burchiel KJ. An analysis of the respective risks of hematoma formation in 361 consecutive morphological and functional stereotactic procedures. Neurosurgery. 2002;50(1): Wang M, Song Z. Improving target registration accuracy in image-guided neurosurgery by optimizing the distribution of fiducial points. Int J Med Robot. 2009;5(1): Shamir RR, Freiman M, Joskowicz L, Spektor S, Shoshan Y. Surface-based facial scan registration in neuronavigation procedures: a clinical study. J Neurosurg. 2009;111(6): Strong EB, Rafii A, Holhweg-Majert B, Fuller SC, Metzger MC. Comparison of 3 optical navigation systems for computer-aided maxillofacial surgery. Arch Otolaryngol Head Neck Surg. 2008;134(10): Pillai P, Sammet S, Ammirati M. Application accuracy of computed tomographybased, image-guided navigation of temporal bone. Neurosurgery. 2008;63(4) (Suppl 2): Greenfield JP, Schwartz TH. Catheter placement for Ommaya reservoirs with frameless surgical navigation: technical note. Stereotact Funct Neurosurg. 2008;86(2): Mascott CR, Sol JC, Bousquet P, Lagarrigue J, Lazorthes Y, Lauwers-Cances V. Quantification of true in vivo (application) accuracy in cranial image-guided surgery: influence of mode of patient registration. Neurosurgery. 2006;59(1)(Suppl 1): ONS146 ONS Mascott CR. In vivo accuracy of image guidance performed using optical tracking and optimized registration. J Neurosurg. 2006;105(4): Mascott CR. Comparison of magnetic tracking and optical tracking by simultaneous use of two independent frameless stereotactic systems. Neurosurgery. 2005;57(4)(Suppl): Holloway KL, Gaede SE, Starr PA, Rosenow JM, Ramakrishnan V, Henderson JM. Frameless stereotaxy using bone fiducial markers for deep brain stimulation. J Neurosurg. 2005;103(3): Ferroli P, Franzini A, Marras C, Maccagnano E, D Incerti L, Broggi G. A simple method to assess accuracy of deep brain stimulation electrode placement: pre-operative stereotactic CT + postoperative MR image fusion. Stereotact Funct Neurosurg. 2004;82(1): 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(4): Steinmeier R, Rachinger J, Kaus M, Ganslandt O, Huk W, Fahlbusch R. Factors influencing the application accuracy of neuronavigation systems. Stereotact Funct Neurosurg. 2000;75(4): Muacevic A, Uhl E, Steiger HJ, Reulen HJ. Accuracy and clinical applicability of a passive marker based frameless neuronavigation system. J Clin Neurosci. 2000;7(5): Dorward NL, Alberti O, Palmer JD, Kitchen ND, Thomas DG. Accuracy of true frameless stereotaxy: in vivo measurement and laboratory phantom studies: technical note. J Neurosurg. 1999;90(1): Vinas FC, Zamorano L, Buciuc R, et al. Application accuracy study of a semipermanent fiducial system for frameless stereotaxis. Comput Aided Surg. 1997;2(5): Maurer CR Jr, Fitzpatrick JM, Wang MY, Galloway RL Jr, Maciunas RJ, Allen GS. Registration of head volume images using implantable fiducial markers. IEEE Trans Med Imaging. 1997;16(4): Nauta HJ. Error assessment during image guided and imaging interactive stereotactic surgery. Comput Med Imaging Graph. 1994;18(4): Maciunas RJ, Galloway RL Jr, Latimer JW. The application accuracy of stereotactic frames. Neurosurgery. 1994;35(4): Maciunas RJ, Galloway RL Jr, Latimer J, et al. An independent application accuracy evaluation of stereotactic frame systems. Stereotact Funct Neurosurg. 1992;58(1-4): Rosenow JM, Sootsman WK. Application accuracy of an electromagnetic fieldbased image-guided navigation system. Stereotact Funct Neurosurg. 2007;85 (2-3): Yaniv Z. Rigid registration. In: Peters T, Cleary K, eds. Image-Guided Interventions: Technology and Applications. New York, NY: Springer; 2008: Fitzpatrick JM, West JB, Maurer CR Jr. Predicting error in rigid-body point-based registration. IEEE Trans Med Imaging. 1998;17(5): Simpson AL, Ma B, Chen EC, Ellis RE, Stewart AJ. Using registration uncertainty visualization in a user study of a simple surgical task. Med Image Comput Comput Assist Interv Int Conf. 2006;9(pt 2): Elder JB, Chen TC. Surgical interventions for primary central nervous system lymphoma. Neurosurg Focus. 2006;21(5):E Sato M, Iwatsuki K, Akiyama C, Kumura E, Yoshimine T. Implantation of a reservoir for refractory chronic subdural hematoma. Neurosurgery. 2001;48(6): Sandberg DI, Bilsky MH, Souweidane MM, Bzdil J, Gutin PH. Ommaya reservoirs for the treatment of leptomeningeal metastases. Neurosurgery. 2000;47(1): Joskowicz L. Advances in image-guided targeting for keyhole neurosurgery: a survey paper. In: McKenna J, Sancha A, eds. US Surgery Touch Briefings; Wadley J, Kitchen N, Thomas D. Image-guided neurosurgery. Hosp Med. 1999;60(1): Vannier MW, Haller JW. Navigation in diagnosis and therapy. Eur J Radiol. 1999;31(2): Galloway RL Jr, Maciunas RJ, Edwards CA II. Interactive image-guided neurosurgery. IEEE Trans Biomed Eng. 1992;39(12): Fitzpatrick JM. Fiducial registration error and target registration error are uncorrelated. Paper presented at: SPIE Medical Imaging; February 7-12, 2009; Orlando, FL. 41. Shamir RR, Joskowicz L. Worst-case analysis of target localization errors in fiducial-based rigid body registration. Paper presented at: SPIE Medical Imaging; February 7-12, 2009; Orlando, FL. 42. Toma AK, Camp S, Watkins LD, Grieve J, Kitchen ND. External ventricular drain insertion accuracy: is there a need for change in practice? Neurosurgery. 2009;65(6): Shamir RR, Joskowicz L, Shoshan Y. Optimal landmarks selection and fiducial marker placement for minimal target registration error in image-guided neurosurgery. Paper presented at: SPIE Medical Imaging; February 7-12, 2009; Orlando, FL. 44. Balachandran R, Fitzpatrick JM. Iterative solution for rigid-body point-based registration with anisotropic weighting. Paper presented at: SPIE Medical Imaging; February 7-12, 2009; Orlando, FL. 45. Moghari MH, Abolmaesumi P. Point-based rigid-body registration using an unscented Kalman filter. IEEE Trans Med Imaging. 2007;26(12): Moghari MH, Abolmaesumi P. Distribution of target registration error for anisotropic and inhomogeneous fiducial localization error. IEEE Trans Med Imaging. 2009;28(6): Wiles AD, Likholyot A, Frantz DD, Peters TM. A statistical model for point-based target registration error with anisotropic fiducial localizer error. IEEE Trans Med Imaging. 2008;27(3): Shamir RR, Joskowicz L, Antiga L, Foroni RI, Shoshan Y. Trajectory planning method for reduced patient risk in image-guided neurosurgery: concept and preliminary results. Paper presented at: SPIE Medical Imaging: February 13-18, 2010; San Diego, CA. COMMENT The authors have demonstrated that frameless neuronavigation is good, but not perfect, in helping to deliver a catheter to its target. This study makes clear that there are many sources for error in NEUROSURGERY VOLUME 68 OPERATIVE NEUROSURGERY 1 MARCH 2011 ons101

8 SHAMIR ET AL neuronavigation and it is very difficult to minimize them all. The authors correctly point out that neurosurgeons should not feel safe just because the navigation system reports a low value for the fiducial registration. Additional care must be taken during placement of flexible devices such as ventricular catheters that can deviate along the planned trajectory. Electromagnetic field based neuronavigation systems use tracked stylets that provide real-time information on the position of the catheter tip, thus allowing online correction of tip deviation. This has become our preferred method, when possible, for placing flexible devices in the brain and ventricles Joshua Rosenow Chicago, Illinois ons102 VOLUME 68 OPERATIVE NEUROSURGERY 1 MARCH