Navigated pedicle screw placement: Experimental comparison between CT- and 3D fluoroscopy-based techniques

Abstract

Objective: Even with CT-based navigation, the misplacement rate for pedicle screws is reported to be as high as 10%. Using fluoroscopy-based 3D navigation, misplacement rates of 1.7 to 6% occur. The purpose of this study was to compare the accuracy of CT-based and Iso-C-based navigation in an experimental context.

Methods: A foam spine model and the SurgiGATE® navigation system were used. First, a determination of point accuracy measured the difference between the real positions of markers placed on selected vertebrae and their positions as determined by the navigation system. In the verification mode, the pointer is placed exactly on the markers displayed on the monitor screen, and the deviation of the pointer tip and marker is measured in reality using a caliper. Secondly, pedicle accuracy was measured using pre-drilled holes for pedicle screws. A trajectory was planned into the visible hole and the navigated pointer was placed.

Results: The measured accuracy for the markers showed a statistically significant difference between the results with CT and Iso-C navigation for one of six markers placed on the vertebra. Iso-C-based navigation demonstrated a lower mean deviation of 0.5 mm, compared to 1 mm with CT-based navigation. The deviation within the pre-drilled holes was lower when using the Iso-C3D scan. Using Iso-C3D navigation, 76.6% of the measurements showed no deviation at the entrance point, compared with 43% when using CT-based navigation. Also, with Iso-C3D navigation, 78.3% of the inserted pedicle awls hit the defined trajectories in the pre-drilled holes correctly, compared to 66.6% with CT-based navigation.

Conclusion: The overall image-to-reality accuracy for CT- and Iso-C-based navigation was assessed in the described experimental setup. An apparent tendency towards higher accuracy with Iso-C-based navigation was evaluated; however, the differences were not significant.

• Iso-C3D navigation
• pedicle screw
• accuracy analysis

Introduction

Currently, transpedicular screw placement is the most accepted technique for posterior stabilization in the thoracolumbar spine [1–3]. Misplacement of these screws may lead to serious vascular or neurological complications [4]. Using conventional techniques, misplacement of screws is reported in up to 30% of cases in the lumbar spine [5],[6] and in up to 55% of cases in the thoracic spine with its thinner pedicles [7]. The incidence of neurological complications arising from such misplacements is reported to be as high as 5% in larger series [4],[8]. In the more recent literature, the incidence of pedicle screw misplacement was reported to be lower in larger series [8],[9], but may actually be higher in centers with less experience.

Surgical navigation for spinal surgery was introduced to improve precision and decrease this misplacement rate [10–16]. The advantage of CT-based navigation is its three-dimensional (3D) visualization of the pedicle, allowing observation of the placement in relation to the anatomical structures. One disadvantage is the need for intraoperative registration and preoperative CT scanning. Newly developed image intensifiers facilitate intraoperative 3D image acquisition [17–19]. Using this new technique in combination with navigation, the registration process can be automated. However, even with the use of CT-based navigation, the misplacement rate in recent clinical trials is reported to be as high as 4.5 to 10% [20–24].

In comparative studies using fluoroscopy-based 3D navigation, misplacement rates of 1.7 to 6% have been reported [24–26], though all cited studies were limited to a small number of clinical cases. To our knowledge, a comparative analysis between the two navigated techniques has not yet been described. Therefore, the purpose of this study was to compare the accuracy of CT-based and Iso-C-based navigation in an experimental context. The goal was to determine whether one of the tested modalities shows greater accuracy than the other, which might have an impact on further clinical applications.

Material and methods

For the experiments, an intact foam model of the entire spine from C1 to the sacrum (Model 9333, Synbone™, Malans, Switzerland) was used, in conjunction with the SurgiGATE® navigation system (Medivision, Oberdorf, Switzerland).

To acquire the CT scans, a spiral CT scanner (Volume Zoom, Siemens, Erlangen, Germany) was used with the following protocol: 120 kV; 150 mAs; slice thickness 1.25 mm; table feed 5.5 mm; reconstruction interval 0.6 mm. The Iso-C3D™ (Siemens AG, Erlangen, Germany) was selected as the 3D fluoroscope. The spine module (version 3.1) of the SurgiGATE system was used for CT-based navigation, and the Iso-C module (version 1.0) was used for Iso-C navigation.

Two experimental setups were used to determine the accuracy of the two modalities:

1. Point accuracy

This setup analyzed the accuracy in terms of the difference between the position of points in reality and in the navigation system. The vertebrae were marked with markers manufactured from titanium k-wire, with a diameter of 1.6 mm and a length of 8 mm, at levels Th4, Th8, Th12, L2, and L4. The markers were placed on the lateral side of the pedicle, and on the lateral and ventral sides of the inferior and superior edges of the vertebra (Figure 1).

Figure 1. Small titanium markers, 1.6 mm in diameter and 8 mm in length, were placed in the lateral pedicle, lateral vertebra and anterior vertebra for the point accuracy test.

A CT scan was performed with the spine in the supine position to simulate normal patient positioning. The marked vertebra was scanned together with the one above and the two below due to the overlapping anatomy of the spinal process in the thoracic spine [27].

The reference base was placed on each individual marked vertebra. Registration was performed using five defined landmarks, as planned in the preoperative mode: the tip of the spinous process and the superior and inferior facets on both sides. Once the matching had been calculated, the quality of the registration was displayed. Using the Calculate/Skip worst button, the system recalculated the matching using one less pair–the worst one. The new matching result was then displayed. If this matching result was larger than 1, the pair-point matching was repeated.

Surface matching was performed when the result of the calculation was good. It was accomplished using 12 points arranged symmetrically on the dorsal vertebra, including the spinal process. Again, if the matching result was larger than 1, the surface matching was repeated.

For Iso-C navigation, the reference base was also mounted rigidly on the marked vertebra. For each vertebra investigated, a new scan was performed. The isocenter was defined using an AP and a lateral fluoroscopic image, placing the vertebra centrally within these images.

The entire spine was placed in special holders at both ends of the model to ensure that no other metal was in the X-ray beam. The holders were themselves placed on a radiolucent table (Figure 2).

Figure 2. Experimental setup for “reversed verification”. The reference base is placed at the navigated vertebra. The pointer is fixed in a special 3D holding device which allows free movement of the pointer in space. On achieving the correct position of the pointer, the device can be fixed in position with the knob.

In the verification mode, the pointer must be placed at a selected point and its position is then compared with the corresponding point in the dataset. In accordance with the procedure described in reference [28], we used the “reversed verification”. After the registration phase, the pointer was placed in a special holder (Figure 2) which allowed 3D movement. It was moved until the tip of the virtually displayed pointer hit the marker to be analyzed on the navigation system (Figure 3). The holder was then rigidly fixed and the real distance between the marker and the tip of the pointer measured using an electronic caliper (CD-15CP, Mitutoyo, Inc., Aurora, IL) with an accuracy of 0.1 μm (Figure 4). All six titanium markers attached to each marked vertebra were selected as reference points for this reversed verification.

Figure 4. The deviation between the titanium marker and the tip of the pointer in reality was measured with an electronic caliper (CD-15CP, Mitutoyo, Inc., Aurora, IL).

Figure 3. Monitor screenshot: The pointer was directed towards the titanium markers and fixed in this position with the holding device.

2. Pedicle accuracy

Drill-holes for the pedicle screws were made in the same vertebrae as mentioned above, with the navigated instruments (i.e., the pedicle awls) being used to obtain the correct diameter prior to acquiring the dataset. The actual placement was performed manually without navigation. A new CT scan (using the same protocol as before) was acquired for each instrumented vertebra, and also included the vertebra above and the two vertebrae below the instrumented vertebra.

In the preoperative planning mode, trajectories (displayed as red lines) within the canal for the pedicle screw were planned exactly. The diameter of the trajectory was planned as 4 mm (matching the diameter of the pedicle awl). A new registration process was performed using the same landmarks for paired-point matching and surface registration as previously mentioned. The reference base was placed on the spinous process of each prepared vertebra. For Iso-C navigation, the reference base was also placed on the spinous process of the prepared vertebra and the scan performed.

After registration, the pedicle awl was inserted into the prepared drill hole. Due to the prior preparation with the same instrument, the pedicle awl fitted the hole without any room for motion. The awl was displayed on the navigation monitor as a green line. A screenshot was acquired from the monitor and transferred using a floppy disc to a laptop computer (Toshiba Satellite 1110) (Figure 5). Using CorelDraw 7 software (Corel Corporation, Ottawa, Ontario, Canada), the screenshots were analyzed, and the deviation of the displayed pedicle awl within the canal from the planned trajectory was determined. The diameter of the trajectory was used as a reference due to the planned diameter of 4 mm. The deviation at the entry point and the maximum difference between the two lines was measured in the lateral and medial as well as cranial and caudal directions. The angular deviation in each direction was also measured.

Figure 5. Monitor screenshot of the measurements for pedicle accuracy. The trajectory (dark grey) was planned into the prepared holes and the navigated pedicle awl (light grey) was placed into the vertebra. On the screenshot, the deviation with respect to the angle and the deviation of the entrance point were measured.

For both experiments, the analysis at each vertebra was performed three times for all markers with a new registration, based on either CT or a new Iso-C scan. The repetitive analysis was performed to achieve an average result for each vertebra. Particular care was taken to avoid movements of the reference base; if such a movement occurred, the trial was redone.

For statistical analysis, SPSS, version 11.5 (SPSS Inc., Chicago, IL), was used. A pairwise comparison, the Mann Whitney Test, was performed. The null hypothesis was that the two imaging devices (CT and Iso-C) did not differ significantly in precision for spinal navigation. The significance level was set at p < 0.05.

Results

Point accuracy

In terms of the marker positions, overall accuracy was greater with Iso-C-based navigation than with CT-based navigation. For the left side vertebra marker, Iso-C navigation showed a significantly lower deviation than CT-based navigation (p = 0.004). There were no statistically significant differences measured for the other markers.

Iso-C-based navigation provided a lower mean deviation than CT-based navigation for all markers. The mean deviation using CT-based navigation was 1 ± 0.83 mm, including all vertebra markers, while with Iso-C navigation it was 0.5 ± 0.48 mm.

The highest deviation with Iso-C-based navigation was 1.88 mm, compared to 2.8 mm with CT-based navigation. With Iso-C navigation, a correct image, i.e., one that was no different from reality, was registered in almost 50% of measurements, whereas with CT-based navigation this result was only obtained in 25% of measurements. No deviation greater than 2 mm was registered with Iso-C navigation, but such deviations were found in 12% of the CT-based procedures (Figure 6).

Figure 6. The point accuracy experiment showed a higher accuracy for Iso-C3D-based navigation. However, the difference between the two modalities was not statistically significant. [Color version available online.]

Pedicle accuracy

Comparing the two imaging modalities for the pre-drilled holes, the deviation in the Iso-C3D scan was lower than when using CT imaging.

Comparing the deviation at the entry point, the Iso-C3D results showed a maximum deviation of 1.3 mm with a mean value of 0.2 ± 0.32 mm. Using the CT-based module, a deviation of 0.58 ± 0.95 mm could be achieved, with a maximum of 4 mm.

Comparing the maximum deviation of the trajectory and the awl, the mean deviation in CT-based navigation was 0.78 ± 1.32 mm, with a maximum value of 4 mm. In Iso-C3D-based navigation, the maximum deviation was 1.5 mm, with a mean value of 0.23 ± 0.43 mm.

Over 75% of the measurements showed no deviation at the entrance point using Iso-C3D-based navigation, compared to 43% with CT-based navigation. No deviation greater than 2 mm occurred with Iso-C3D (Figure 7).

Figure 7. Measurement of the total deviation (in mm) at the entrance point showed no statistically significant difference between the two navigation modalities. [Color version available online.]

Comparing the deviation of the pedicle axis, 78.3% of the inserted pedicle awls hit the defined trajectories in the pre-drilled holes correctly when using Iso-C3D navigation. Also, less than 5% had an angular deviation of 1–2°, and no deviation greater than 2° occurred with the Iso-C. For CT-based navigation, two thirds of the inserted pedicle awls showed no deviation and a deviation of more than 2° occurred in 6.6%.

The mean deviation for Iso-C-based navigation was 0.2 ± 0.6°, with a maximum value of 1.9°, as compared to a mean deviation of 0.6 ± 1.1° and a maximum of 4.5° for CT-based navigation (Figure 8).

Figure 8. Measurement of the deviation of the pedicle axis in the “reversed mode” showed no statistically significant difference between the two navigation modules. [Color version available online.]

Discussion

The challenges involved in transpedicular screw placement arise from the variations in human spinal anatomy and the difficulties associated with intraoperative recognition of the pedicle arrangement. Misplacement of pedicle screws may lead to serious vascular or neurological complications [4]. In addition, intraoperative patient movement may affect the accuracy of navigated pedicle screw placement [24],[35]. Gebhardt et al. [27] and Arand et al. [29] reported registration errors due to soft tissue or osteoporotic bones. As described by Hott et al. [30], a decrease in image quality in osteopenic or obese patients should be kept in mind in clinical series. Therefore, we used a plastic spine model in our experiments to evaluate the basic accuracy of the two imaging modalities in navigated pedicle screw placement.

In a recent clinical study, Rampersaud et al. [31] found an overall pedicle wall breach incidence of 15.3%. Their pedicle screw placement was performed under fluoroscopic navigation between T2 and S1. On evaluating the misplacement, the incidence in the thoracic spine was found to be significantly greater than that in the lumbar spine: 31.6% versus 10.6%.

In the literature, the higher misplacement rate has been attributed to the smaller pedicles in the thoracic spine, the different orientation (i.e., greater medial orientation) of the pedicles compared to those in the lumbar spine, and the less distinct anatomic references for the starting point [32–35].

In our study, we were unable to detect any difference between the thoracic and lumbar spine due to the special experimental setup using “reversed verification”. However, we used several levels of the spinous model to achieve a representative result, especially for the matching results for CT-based navigation.

The basic accuracy under optimal experimental conditions for CT- and Iso-C3D-based navigation was analyzed. The question to be answered was whether either method provided superior accuracy for spinal navigation. In the described experimental setup concerning the overall image-to-reality accuracy using a “reversed verification” model, we were able to demonstrate greater accuracy for Iso-C-based navigation; however, the differences were not statistically significant.

One drawback of this experimental setup might be the small number of attempts for each vertebra. This could be one reason for the lack of a significant difference in accuracy between the two tested modalities for spinal navigation. Another disadvantage of our study design was that we were unable to report misplacement rates because of the special setup. Furthermore, due to the “reverse verification” approach, an inaccuracy caused by artefact within the scan, either CT or Iso-C3D, may occur, resulting in a mismatch of the pointer on the monitor. However, these errors may occur with either modality and may be similar due to there being only a single observer.

With Iso-C3D-based navigation, the misplacement rate described is lower than that for CT-based and fluoroscopy-based navigation. Grützner et al. [25] reported a misplacement rate of not more than 2 mm in 1.7% of 302 pedicle screws implanted using Iso-C3D navigation. Wendel et al. [24] reported 0.7% misplacement in 141 screws placed with Iso-C-based navigation, as compared to a misplacement rate of 4.5% with CT and 2.8% with fluoroscopy.

Holly and Foley [26] performed a laboratory evaluation of Iso-C navigation using three fresh-frozen intact human torsos. They placed a reference on a spinous process and performed the navigated screw placement at this level and at one vertebra above and below. The pedicles of T1 down to L5 were instrumented. For image guided drill placement, a small skin incision was made. Of 102 pedicle screws placed, 94.7% were placed correctly, with 100% correct placement in the lumbar spine and 92% correct placement in the thoracic spine.

In general, when using Iso-C navigation, a larger deviation can be observed when placing the reference point farther away from the isocenter, so we performed the accuracy analysis by placing the reference on the vertebra being investigated [36].

One reason for the higher accuracy in lumbar spine pedicle screw placement might be the different pedicle anatomy. Rampersaud et al. [37] investigated the “safe” corridor for pedicle screw placement, using the dimensions of pedicles found in several other studies. These data were compiled to determine a mathematical model of the pedicles in cervical, thoracic and lumbar spine. From these parameters, they calculated the maximum tolerable error for translational and rotational tolerances when placing pedicle screws. The highest required accuracy was found to be for Th3 to Th7 and for the thoracolumbar junction from Th11 up to L2. In total, Rampersaud et al. reported a maximum translational error of 1 mm and less than 5° for this region.

When comparing the results of Rampersaud et al. with the deviation found in our own study, we can claim a lower misplacement rate for Iso-C3D navigation, compared with CT-based navigation. With both modalities, the highest deviations were lower than 5°; however, the maximum translational error was up to 4 mm with CT-based navigation and up to 1.3 mm with Iso-C3D-based navigation using plastic models, which give greater accuracy compared to cadaveric specimens in experimental studies [38]. This is one reason for using plastic models in the presented study: to minimize other errors and to evaluate the basic accuracy of the two 3D imaging modalities.

Currently, there are only a few published studies demonstrating the deviation of the planned screw and the achieved pedicle screw drilling.

Nakagawa et al. [39] performed a laboratory study using lumbar plastic models consisting of L1 to L5 and fluoroscopy-based navigation. They performed the drilling with a 3.5-mm drill bit and measured the angular deviation between the planned and executed drilling on lateral and axial images. The mean deviation on the axial images was 1.3° (range: 0–5°) and on the sagittal images 1.0° (range: 0–6°). However, in this study, the lateral deviation of the drill hole was not mentioned.

Foley et al. [40] used an experimental setup with cadaveric specimens and fluoroscopy-based navigation. They performed an accuracy analysis by planning a trajectory for the instrumentation within the acquired lateral image for navigation. After placing the probe into the pedicle, a new lateral image was acquired and the deviations between the planned and achieved angle were measured. They found a mean deviation of 2.7° and a maximum deviation of 5°. The mean probe tip error was 0.97 mm, with a maximum of 3 mm.

Kamimura et al. [41] performed a laboratory accuracy test for CT-based navigation with plastic lumbar spine models and analyzed the deviation between the hole position in the plastic model and the planned screw position. When measuring the deviation in the frontal plane of the pedicle canal, they measured the distance from the center of the pedicle to the hole in this plane. They found a deviation of 1.78 ± 0.81 mm (range: 0.11–3.69 mm). The angular deviation was measured in the axial plane as 2.28 ± 1.92° (range: 0.0–9.0°).

In our experiments, the deviation of the entrance point of the drilling was measured. Comparing the deviation with CT-based navigation and Iso-C-based navigation, we found a lower deviation of the entrance point and angular deviation. However, for CT-based navigation we performed a supplementary surface matching. Therefore, we believe, we were able to achieve a high level of precision using CT-based navigation.

In previous reports concerning the accuracy of pedicle screw placement, there has been no clear indication as to whether the authors used a rigid drill sleeve to help control possible drill bending as reported by previous researchers [42–44]. To avoid this potential error, we used a rigid pedicle screw placement.

In summary, in this experimental study, the overall accuracy of both 3D navigation tools was high; however, the measured point and pedicle placement accuracy was improved when using Iso-C3D-based navigation as compared to CT-based navigation, though not significantly. Furthermore, Iso-C3D-based navigation allows the user to have an immediately accessible intraoperative 3D CT. Future clinical studies will be designed to improve on our current findings.