Until the mid 1990s, pedicle screws generally were inserted using visual and/or fluoroscopic control. Computed tomography (CT)-controlled studies of conventional transpedicle screw placement have shown substantial rates of incorrect positioning in the lumbar spine (28-40%)6,11 and implant-related neurologic complications (> 2%).7,29
Computer-aided systems have been introduced for CT-based22 and C-arm-based13 freehand navigation. With CT-based computer guidance, experimental1,5,23 and clinical18,20,30 studies have indicated an increased accuracy of transpedicle implant placement with perforation rates of 3% to 11%. Comparative studies1,18 have shown greater rates of misplacement in conventionally treated cohorts (14-42%) compared with computer-assisted groups. In a laboratory experiment,13 the accuracy of C-arm-based navigation showed a maximum deviation of an instrument to the position of a defined steel ball of 0.2 mm. The in vivo investigations resulted in perforation rates of 5.6% to 17%.9,21,25
In a morphometric study, Rampersaud et al26 described very small corridors for safe placement of intraosseous transpedicle 5-mm screws in the thoracic spine (minimum, 0 mm translational and 0° rotational error in T5; maximum, 1.5 mm translational and 7.7° rotational error in T1) and 6.5-mm screws in the lumbar spine (minimum, 0.65 mm translational and 2.1° rotational error in L1; maximum, 3.8 mm translational and 12° rotational error in L5).
Despite various experimental and clinical investigations with qualitative analyses of the percentage of pedicle breaches, no studies have examined the quantitative accuracy of computer-assisted orthopaedic surgery (CAOS) in standardized thoracic and lumbar spine models, and none has compared the accuracy of CT-based navigation and C-arm-based navigation at different locations. We think it is crucial to know the accuracy required for safe screw placement,26 and it is important to state a navigation system's precision capabilities to assess risk.
Because of the frequency of pedicle perforations in various navigated cohorts,2,3,20,27 we hypothesized that submillimeter accuracy, as previously suggested,21 is not realistic with CAOS when using it for transpedicle instrumentation. Based on our previous clinical observations,2 we also hypothesized that CT-based navigation in three-dimensional (3D) datasets may lead to more accurate results than C-arm-based navigation in two-dimensional (2D) datasets, considering the different uses of navigation at the thoracic and lumbar spine.2,25 Finally, we hypothesized that the incidence of pedicle perforation (qualitative accuracy) may not sufficiently verify the accuracy of surgical targeting (quantitative accuracy) and therefore is not a reliable measure in characterizing precision of CAOS in transpedicle spinal instrumentation.
MATERIALS AND METHODS
To evaluate the quantitative accuracies of conventional guidance and CT-based and C-arm-based navigation, we created a model to judge precision based on plastic thoracic and lumbar vertebrae (Fig 1). We studied 90 artificial spine vertebrae (Sawbone©, Limhamn, Sweden) (30 per group) by conventional guidance, CT-based navigation, and C-arm-based navigation. We used three vertebrae each from T6-T10 and L1-L5. As a defined target for the 3.2-mm drill hole in each vertebra, a 4-mm steel ball was fixed just anterior to the base of the extension of the left pedicle. The ball was fixed with instant adhesive to prevent it from moving during drilling (Uhu plus, Endfest 300©, Bühl/Baden, Germany). In the two navigated techniques, the virtual 3.2-mm canal was guided precisely into the center of the steel ball before drilling, while supported by a frame. With the conventional technique, the same drillings were completed by freehand guidance using the drilling machine under fluoroscopic control. The accuracy of the drillings was determined by CT evaluation and vector calculation.
We used an optoelectronic navigation system (SurgiGATE©, Praxim, La Tronche, France). The navigable datasets for the CT-based 3D navigation were generated on a four-gantry spiral CT (MX 8000©, Marconi, Hofheim-Wallau, Germany). For the navigated and conventional 2D datasets, we used an Iso-C-arm (Siremobil Iso-C©, Siemens AG, Erlangen, Germany) with an isocentric center of rotation.
One transpedicle hole was drilled in each vertebra via the left pedicle by the senior author (MA), using one of the three guidance techniques. The goal was to precisely target the drill canal onto the center of the steel ball without perforating the left pedicle wall.
With the conventional technique, the specimen was fixed to allow anteroposterior (AP) and lateral fluoroscopic views. An exact left transpedicle AP view was adjusted with the image intensifier so the steel ball was seen in the center of the pedicle. A small pedicle awl with a transparent plastic handle was placed into the posterior 5 mm of the left pedicle using fluoroscopy, superimposed, and aimed toward the steel ball. The pedicle awl was removed during constant lateral fluoroscopic control. Maintaining its direction, a 3.2-mm drill with a power drive (Kolibri®, Synthes, Oberndorf, Switzerland) was directed to the center of the steel ball. Each step of the procedure was documented with screen-shots.
In the CT and C-arm navigated procedures, the holes were drilled by a navigated drilling machine (Kolibri®, Synthes, Oberndorf, Switzerland) fixed to a frame. This only allowed the system to move in the direction of the drill (Fig 2). To minimize bending of the drill during use, we used a specially stiffened drill with a strengthened proximal 6-mm shaft; we used the same drill clinically.
In the CT-based navigation group, the standardized navigable CT datasets of all 30 vertebrae were collected in a plastic frame, transferred onto the navigation system, followed by planning of the matching procedure and the trajectory. Because all data were collected using the same CT protocol, the same threshold was used for every specimen. Equivalent landmarks to clinical conditions were collected. We determined the most cranial and caudal corner of the spinous process in the lumbar spine and the most posterior extension of the mammillary processes (left and right). In the thoracic spine, spinous process landmarks and the center of the mediolateral ledge of the left and right articular masses were added. Additionally, the trajectory in the left transpedicle axis was planned by targeting exactly into the center of the steel ball without perforating the transpedicle bone. During the experiment, the corresponding dataset of each vertebra was loaded into the navigation system, and a dynamic reference base was rigidly fixed to the spinous process. The matching procedure consisted of paired-point matching with the four defined landmarks and surface matching with 12 points distributed over the posterior aspect of the vertebra. Accuracy of the matching procedure was verified by contacting different posterior points on the vertebra with a pointer. The experiment was done only in specimens with verified precise matching, otherwise rematching was done. In the next step, the correct navigated alignment of the extension of the virtual drill canal and the center of the steel ball was adjusted. The vertebra was held by a socket joint. It was rigidly fixed after achieving the desired complete correspondence between the planned trajectory and guided instrumentation (Fig 3). After taking a screenshot, we drilled by moving the drilling frame forward exactly along the axis of the drill. This was accomplished without influencing the direction of the drill by any guidance. Drilling was done just until the steel ball was reached, as indicated by the navigation system. Another screenshot showed the navigated situation after drilling (Fig 4).
In the same experimental setup, the C-arm navigated target drillings were investigated in 30 vertebrae. Using this technique, the navigable C-arm datasets were collected in corresponding fluoroscopic planes as described for the conventional technique (transpedicle AP view with the steel ball in the center of the observed pedicle and lateral view), under constant tracking of the image intensifier, dynamic reference base, and gravitation base. Next, the specimen was positioned in the drilling frame with the aimed instrument on the navigation monitor superimposed on the virtual drilling canal to the steel ball (Fig 5). Then the left trans-pedicle hole was drilled without using computer guidance. After the drilling procedure, the status of the navigation system was documented with a screenshot (Fig 6).
To evaluate the quantitative accuracy of the three guidance techniques, standardized thin-slice frontal-plain CT (0.6 mm) was done for each specimen. Using graphics software (Osiris©, NIH, Washington, DC), we determined the orientation of the drill canal relative to the steel ball (Fig 7). Therefore, the coordinates of the center of the start and end points of the drilled hole and the center of the steel ball were identified on defined CT slices. The shortest distance between the straight line defined by the center of the start point and the end point of the drill hole and the center of the ball was calculated via vector calculation (quantitative accuracy). Therefore, the virtual straight line through the center of the drill canal can be described as:
The constant λ refers to the point of the straight line located next to the center of the steel ball:
With the constant λ, the coordinates of the intersection between the perpendicular to the center of the ball and the straight line can be determined (xv, yv, and zv) to evaluate the direction of misplacement of the drilled canal in relation to the target.
Direct calculation of the minimum (perpendicular) distance between the center of the straight line through the drill canal and the center of the steel ball is possible using this equation:
The calculations were performed using Excel® (Microsoft Corporation, Bellevue, WA). A transformation factor caused by the image acquisition of the CT in pixel values had to be used to convert the data into metric measurements.
To evaluate qualitative accuracy, all specimens were visually inspected by two independent observers (TF, GK), and all pedicle perforations were noted. When a pedicle breach was found, a 3.2-mm steel body was inserted into the drill hole. The distance of the perforation in relation to the outside wall of the pedicle was measured mechanically.
Statistical analysis was done using SAS® (Statistical Analysis System, SAS, Cary, NC). To investigate differences between the three guidance techniques in the thoracic and lumbar vertebrae, we evaluated minimum distance between the straight line defined by the center of the start point and the end point of the drill hole and the center of the steel ball, 25% quartile, median, 75% quartile, and maximum distances.
The 95% confidence interval of each median was calculated for all three guidance techniques. Without overlapping the results in two groups, a high probability of discrimination of the real medians was assumed. We used a Kruskal-Wallis test to examine whether the minimum distances of the drill canal to the steel ball for the three guidance techniques were different. Statistical significance was set at p < 0.10. If this test suggested a tendency, a Wilcoxon Mann-Whitney U test was used to calculate differences between two guidance techniques. Statistical significance was set at p < 0.05. We used only nonparametric statistical tests because the data were not normally distributed.
The vector calculation (quantitative accuracy) showed that submillimeter accuracy of CAOS in the transpedicle standardized drilling model was not consistently achievable using navigation (Figs. 8, 9). With CT-based computer guidance, seven of 30 drill holes were located less than 1 mm from the steel ball; the median of accuracy was 1.6 mm. With C-arm navigation, only one of 30 drillings had a distance less than 1 mm from the goal; the median precision was 2.3 mm.
Computed tomography-based navigation achieved the most accurate results, followed by C-arm based navigation and the conventional technique. Computed tomography-based navigation also showed the highest accuracy in the thoracic specimens, whereas C-arm navigation had the most precise results in the lumbar spine. The conventional technique showed a median of 2.2 mm (first to third quartiles, 1.4-3.15 mm; range, 0.5-4 mm) in the thoracic specimen. Computed tomography-based navigation led to increased (p = 0.1113) accuracy with a 1.4-mm median deviation (first to third quartiles, 1.2-1.7 mm; range, 0.5- 4.8 mm) of the canal center from the center of the steel ball. In contrast, the C-arm navigated technique showed more inaccuracy (p = 0.7734) than the conventional technique with a median of 2.6 mm (first to third quartiles, 1.7-3.2 mm; range, 0.9-4.8 mm). The comparison between CT-based and C-arm-based navigation confirmed the CT-based technique was more accurate (p = 0.0426), but not substantially so. In the lumbar specimen, the conventional technique resulted in a 2.7-mm median deviation (first to third quartiles, 2.35-4.35 mm; range, 1.3-5 mm) of the canal center from the center of the steel ball. Computed tomography-based navigation showed increased (p = 0.0036) accuracy with a median of 1.8 mm (first to third quartiles, 1.2-2.3 mm; range, 0.5-3 mm). Compared with the conventional technique, C-arm navigation showed a more precise (p = 0.0084) performance with a median of 2 mm (first to third quartiles, 1.8-2.35 mm; range, 1.2-3 mm). We observed no difference in accuracy between CT-based and C-arm-based navigation.
The incidence of pedicle perforation of drill holes verified by visual inspection (qualitative accuracy, Table 1) showed no correlation with the targeting accuracy on the steel ball determined by vector calculation (quantitative accuracy). We categorized the two visual inspection groups as perforation and nonperforation groups. For the perforation group, the vector-calculated median of the canal center from the center of the steel ball was 2.5 mm (95% confidence interval [CI], 1.4-4 mm), and for the nonperforation group it was 2.2 mm (95% CI, 0.5-3.3 mm). Because of the relevant overlapping of the 95% confidence intervals, we assumed there was no difference in the vector-calculated medians of the canal center from the center of the steel ball between groups.
We intended to determine the safety of CAOS transpedicle drillings. Although experimental5 and clinical15 series with 0% pedicle breaches have been reported, we hypothesized that the accuracy of computer guidance at the spine is overestimated. We investigated the precision of conventional, CT-based, and C-arm-based navigated 3.2-mm transpedicle drills toward a 4-mm steel ball in a standardized thoracic and lumbar spine setup, using a drilling frame for the navigated techniques to avoid any influence on the position of the drill canal in the pedicle.
Despite advantages of the chosen drill model concerning reproduction, standardization, and CT evaluation versus a screw placement model is a systematic limitation of the study and a source of inaccuracy because of the movements between the drill and the tracked power drive in navigated guidance. The summed coordinates of the center of all drill holes showed a slight cranial shift on the y axis. This translocation for both navigated techniques can be estimated up to 1 mm per each canal because of bending moments on the drill bit while drilling. All components in the study, including the navigated power drive and the drill bit, are instruments used intraoperatively, therefore a realistic and not a preferable situation for computer-guided results was created. Additionally, the shift was identical in both navigated modalities, so this systematic fault does not jeopardize the comparison of the navigated groups. The frame-assisted drilling in the computer-guided groups excluded observer influence on the results versus freehand drilling in the conventional technique. Preliminary experiments showed AP view radiographs of the transpedicle, essential for precise conventional targeting, could not be achieved in the drill frame because of superimposition of metal. Because of the different drilling techniques, the comparison of the conventional and navigated results should be considered with reservation.
In studies examining the accuracy of CT-based and C-arm-based navigation13,22,24 (Table 2), the drill seems to be the most sensitive point (reduced by an extended 6-mm shaft in this study).14
Few attempts have been made to quantitatively determine the accuracy of navigated lumbar transpedicle screw placement.4,5,8,15 Kamimura et al15 used a transpedicle drilling model to determine the accuracy of CT-based navigation using the freehand technique. No perforation of the lumbar pedicle bone in 44 bilateral drilled pedicles occurred as confirmed by outside inspection of the samples.15 They reported a CT-guided technique precision of 1.78 mm ± 0.81 mm.15 In a nonstandardized CT-based navigation model, a precision of 1.4 ± 0.6 mm for the tip of the Schanz screw in 44 plastic lumbar pedicles was reported.5 The same guiding technique used on human lumbar pedicles resulted in an accuracy of 1.6 ± 0.75 mm.5 In another CT-based navigated human lumbar model, the accuracy from L1 to L5 was 1 to 1.9 mm, with a range between 0.4-3.4 mm.4 The accuracy after reaming the pedicles of 12 human lumbar vertebrae using fluoroscopic navigation was investigated in only one study.8 Foley et al calculated a precision of 0.97 mm ± 0.4 mm. The maximum inaccuracy was 3 mm.8 The findings of all accuracy studies investigating CT-based and C-arm-based navigation were comparable with our in vivo study results.
The first hypothesis, that submillimeter accuracy cannot be provided by a navigation system, is supported by data from the literature and our results. Therefore, an important question of our study was whether the navigation system was safe to use in the thoracic and lumbar spine. Because we had a defined end point for the drilling but variable starting points, we assessed only translational errors at the steel ball and not rotational errors of the drill canal, as previously reported.26 The x axis deviation from the target (medial and lateral) was calculated for every drill canal (Table 3). Based on our accuracy results, CT-based navigation is only safe from L3-L5 and C-arm-based navigation is safe from L2-L5. In the T6, T8-T10, and L1 levels for both navigated guidance techniques, and L2 for CT-based navigation, the precision may be accurate but it is not safe. Therefore, a perforation of the pedicle can occur even under optimal conditions. For T7, neither the navigated nor the fluoroscopic guidance is safe. These results are supported by how pedicle perforations in the thoracic spine2,27 increased compared with the lumbar spine2,17 in our study and in clinical studies.
Our results also showed CT-based navigation in the thoracic spine was more accurate than C-arm navigation, although no difference between these two guidance modalities was present at the lumbar spine. Comparative experimental investigations on the spine have not been published previously, but some clinical reports suggest a comparable incidence of perforations in the thoracic9 and lumbar spine after transpedicle instrumentation.2 Our data support these findings.
The accuracy of transpedicle screw placement, described as successful when a complete intraosseous position was provided9,12,18,20,28 is not an adequate indicator in terms of investigating the precision capabilities of a navigation system. We showed that a pedicle breach of the drill canal does not correlate with an inaccurate result in targeting the steel ball.
We evaluated the accuracy of CT-based and C-arm- based navigation in a standardized in vitro setup. We found the CT-based and C-arm-based computer guidance techniques were superior. However, only the CT-based navigation technique showed increased accuracy when used in the thoracic spine. We think clinical use of C-arm navigation is safe in the lumbar spine. Submillimeter accuracy was not achievable. The inaccuracies resulted from internal and external factors of the navigation system, particularly from compounded instruments like the navigated power drive and drill.
We thank Dr. Michael J. Broom, Florida Spinecare Center, Orlando, for editing the manuscript.
1. Amiot LP, Lang K, Putzier M, Zippel H, Labelle H. Comparative results between conventional and computer-assisted pedicle screw Clinical Orthopaedics and Related Research installation in the thoracic, lumbar, and sacral spine. Spine
2. Arand M, Hartwig E, Kinzl L, Gebhard F. Spinal navigation in tumor surgery of the thoracic spine: first clinical results. Clin Orthop Relat Res
3. Arand M, Schempf M, Hebold D, Teller S, Kinzl L, Gebhard F. Precision of navigation-assisted surgery of the thoracic and lumbar spine. Unfallchir
4. Berlemann U, Monin D, Arm E, Nolte LP, Ozdoba C. Planning and insertion of pedicle screws with computer assistance. J Spinal Disord
5. Carl AL, Khanuja HS, Sachs BL, Gatto CA, vomLehn J, Vosburgh K, Schenck J, Lorensen W, Rohling K, Disler D. In vitro simulation: early results of stereotaxy for pedicle screw placement. Spine
6. Castro WH, Halm H, Jerosch J, Malms J, Steinbeck J, Blasius S. Accuracy of pedicle screw placement in lumbar vertebrae. Spine
7. Esses SI, Sachs BL, Dreyzin V. Complications associated with the technique of pedicle screw fixation: a selected survey of ABS members. Spine
8. Foley KT, Simon DA, Rampersaud YR. Virtual fluoroscopy: computer-assisted fluoroscopic navigation. Spine
9. Fritsch E, Duchow J, Seil R, Grunwald I, Reith W. Accuracy of fluoroscopic navigation of pedicle screws: CT-based evaluation of bone screw placement. Orthopade
10. Gebhard F, Weidner A, Liener UC, Stöckle U, Arand M. Navigation at the spine. Injury
. 2004;35(suppl 1):S-A35-45.
11. Gertzbein SD, Robbins SE. Accuracy of pedicular screw placement in vivo. Spine
12. Girardi FP, Cammisa FP, Sandhu HS, Alvarez L. The placement of lumbar pedicle screws using computerised stereotactic guidance. J Bone Joint Surg Br. 1999;81:825-829
13. Hofstetter R, Slomczykowski M, Sati M, Nolte LP. Fluoroscopy as an imaging means for computer-assisted surgical navigation. Comput Aided Surg
14. Hüfner T, Geerling J, Oldag G, Richter M, Kfuri M, Pohlemann T, Krettek C. Accuracy study of computer-assisted drilling: the effect of bone density, drill bit characteristics, and use of a mechanical guide. J Orthop Trauma
15. Kamimura M, Ebara S, Itoh H, Tateiwa Y, Kinoshita T, Takaoka K. Accurate pedicle screw insertion under the control of a computer-assisted image guiding system: laboratory test and clinical study. J Orthop Sci
16. Khadem R, Yeh CC, Sadeghi-Tehrani M, Bax MR, Johnson JA, Welch JN, Wilkinson EP, Shahidi R. Comparative tracking error analysis of five different optical tracking systems. Comput Aided Surg
17. Laine T, Lund T, Ylikoski M, Lohikoski J, Schlenzka D. Accuracy of pedicle screw insertion with and without computer assistance: a randomized controlled clinical study in 100 consecutive patients. Eur Spine J
18. Laine T, Schlenzka D, Makitalo K, Tallroth K, Nolte LP, Visarius H. Improved accuracy of pedicle screw insertion with computer-assisted surgery: a prospective clinical trial of 30 patients. Spine
19. Marmulla R, Hilbert M, Niederdellmann H. Inherent precision of mechanical, infrared and laser-guided navigation systems for computer-assisted surgery. J Craniomaxillofac Surg
20. Merloz P, Tonetti J, Pittet L, Coulomb M, Lavallee S, Troccaz J, Cinquin P, Sautot P. Computer-assisted spine surgery. Comput Aided Surg
21. Nolte LP, Slomczykowski MA, Berlemann U, Strauss MJ, Hofstetter R, Schlenzka D, Laine T, Lund T. A new approach to computer-aided spine surgery: fluoroscopy-based surgical navigation. Eur Spine J
22. Nolte LP, Visarius H, Arm E, Langlotz F, Schwarzenbach O, Zamorano L. Computer-aided fixation of spinal implants. J Image Guid Surg
23. Nolte LP, Zamorano LJ, Jiang Z, Wang Q, Langlotz F, Berlemann U. Image-guided insertion of transpedicular screws: a laboratory set-up. Spine
24. Nolte LP, Zamorano L, Visarius H, Berlemann U, Langlotz F, Arm E, Schwarzenbach O. Clinical evaluation of a system for precision enhancement in spine surgery. Clin Biomech (Bristol, Avon)
25. Rampersaud YR, Pik JH, Salonen D, Farooq S. Clinical accuracy of fluoroscopic computer-assisted pedicle screw fixation: a CT analysis. Spine
26. Rampersaud YR, Simon DA, Foley KT. Accuracy requirements for image-guided spinal pedicle screw placement. Spine
27. Schnake KJ, Konig B, Berth U, Schroeder RJ, Kandziora F, Stockle U, Raschke M, Haas NP. Accuracy of CT-based navigation of pedicle screws in the thoracic spine compared with conventional technique. Unfallchir
28. Schwarzenbach O, Berlemann U, Jost B, Visarius H, Arm E, Langlotz F, Nolte LP, Ozdoba C. Accuracy of computer-assisted pedicle screw placement: an in vivo computed tomography analysis. Spine
29. West JL3rd, Bradford DS, Ogilvie JW. Results of spinal arthrodesis with pedicle screw-plate fixation. J Bone Joint Surg Am
. 1991;73: 1179-1184.
30. Youkilis AS, Quint DJ, McGillicuddy JE, Papadopoulos SM. Stereotactic navigation for placement of pedicle screws in the thoracic spine. Neurosurgery
© 2006 Lippincott Williams & Wilkins, Inc.
31. Zamorano LJ, Nolte LP, Kadi AM, Jiang Z. Interactive intraoperative localization using an infrared-based system. Neurol Res