Robot arm based flat panel CT-guided electromagnetic tracked spine interventions: phantom and animal model experiments

Introduction
The accurate and timely delivery of diagnostic and therapeutic material in an image-guided procedure is crucial for the successful completion of such procedures. Constant monitoring of the position of the instrument in relation to the patients anatomy ensures exact placement of the interventional device. Spinal intervention such as facet joint (FJ) infiltration or intervertebral disc (IVD) punctures are frequently carried out using standard C-arm X-ray appliances (CAXA) or computer tomography (CT).

While CAXA is widely available, it provides projectional imaging in only one plane, leading to multiple repositioning steps during interventions. CT provides high spatial resolution cross-sectional imaging at a good bone to soft tissue contrast^1,2. Both techniques have the disadvantage of radiation exposure during the tracking procedure especially if the needle needs to be tracked during advancement or repositioning. Trajectory assistance systems, e.g. using laser light to visualise needle insertion points and angles, facilitate the insertion of interventional instruments but provide no information about the current instrument tip position^1,3.

Tracking and navigation systems have been introduced into clinical routine to allow for constant tracking of instrument positions during interventions. Working either by triangulating infrared markers attached to the needle’s end (optical tracking) or measuring the induced current in small sensor coils integrated in the instrument’s tip when placed in an electromagnetic field (electromagnetic tracking, EMT) no additional radiation is needed during the tracking procedure, reducing radiation exposure to the interventionalist and the patient^4,5.

Optical tracking provides robust tracking with the drawback of only indirectly tracking the instrument’s tip by calculating it from the rear mounted markers. Although electromagnetic tracking directly monitors the instrument’s tip, it can create electro- magnetic noise because of ferromagnetic material in the interventional field.

Another recent development in the field of medical imaging is the combination of robotic arm mounted flat panel (RMFP) imaging systems. Such systems combine the possibility of 2D projectional imaging with the capability of large volume 3D cross-sectional CT-like imaging^6,7. RMFP systems are designed as on-demand imaging appliances featuring swift parking of the systems to allow for full patient access. The appliances are equipped with electronically controllable carbon fibre patient supports to allow for reproducible positioning and overlay free imaging.

In this study an electromagnetical tracking system was used in combination with a robotic arm flat detector system. This combination seemed beneficial as the advantage of large volume imaging is combined with direct (tip tracked) navigation in an environment without electromagnetical or ferromagnetic disturbance. The goal of this study was to assess spatial accuracy and procedure times by employing the system in a clinical environment using phantom and animal cadaveric models.

Material and Methods
For all procedures a robot mounted flat panel detector-based imaging system (Artis zeego, Siemens Healthcare, Forchheim, Germany) was used in combination with a clinically approved electromagnetic tracking system (iGuide CAPPA, CAS innovations, Erlangen, Germany).

Figure 1. Movement degrees of freedom for the RMFP (red arrows). In total the robot has six rotational and one translational degree of freedom

Robot arm imaging and electromagnetic tracking system
The RMFP unit consists of a high performance C-arm flat panel detector system mounted on an industrial robot arm. The robotic arm allows for fast, exact and reproducible placement of the detector system in relation to the patient. The wide movement range (Fig. 1) of the robotic arm allows for variable placement of the C-arm around the patient including on demand functionality by swiftly parking the C-arm outside the interventional field (Fig. 2).

Figure 2. Movement during image acquisition (frames 1–3) and parking of the RMFP (frames 4–6), the setup as in the animal cadaveric experiments. The imaging rotation describes a slightly eccentric elliptic path which is performed twice in the case of large volume acquisition. One imaging rotation takes around 5 s, two rotations (large volume scan) with intermittent position adjustment 15 s and parking the detector around 15 s.

In addition to projectional X-ray imaging using the 30-40 cm detector, the system is able – by precisely performing complex movement patterns with the RMFP system – to acquire large volume (40-40-24 cm) CT-like images, outperforming standard C-arm based 3D volume imaging (Fig. 4).

Figure 3. Phantom experiment setup (RA robot arm, parked; FD flat panel detector; CF carbon fibre table; CU electromagnetic tracking system control unit; FG field generator; RP reference plate).

For overlay free imaging the system is used in combination with an electronically controlled carbon fibre based patient support (Fig. 3).

A patient use approved navigation system was used to perform the interventions (iGuide CAPPA, Siemens Healthcare, Forchheim, Germany). The system consists of an electromagnetic tracking system (NDI Aurora, Northern Digital, Waterloo, Ontario, Canada) comprising an arm-mounted field generator unit and a system control unit, as well as a PC-based system running the puncture- controlling application. For referencing purpose a registration plate is used featuring five radio-opaque marker spheres and a six degrees-of-freedom (DOF) electromagnetic sensor coil to match the EMT coordinate system to the imaging information.

Figure 4. Rotationally acquired projectional imaging (upper row) is converted into CT-like images (lower row) with a field of view of ca. 35 _ 35 _ 24 cm (coronal _ sagittal _ axial).

Phantom experiments
An anthropomorphic lumbar spine model (Sawbones, Pacific Research Laboratories, Vashon, Washington, USA) embedded in gel wax (Fig. 3) was used to simulate an anatomical soft tissue covered lower back model. The model was placed in prone position on the RMFP patient support with the reference plate fixed by using transparent draping. Large volume planning RMFP imaging was performed (2 _ 273 projectional images, 90 kV tube voltage, 68 mAs tube current, 0.791 ms exposure time/image) and axial reformations were rendered (0.89 mm³ voxel size, Fig. 4) and transferred to the EMT system by using standard DICOM transfer.

Direct and angulated punctures of the facet joints on both sides as well as punctures of the intervertebral discs were planned. Subsequently the punctures were carried out using an 18-G electromagnetically tracked puncture needle (CAS innovations, Siemens Healthcare, Erlangen, Germany, Fig. 5).

Figure 5. Electromagnetically tracked needle as used in the study. The needle consists of a sheath (white arrow) with a removable mandrin. The electromagnetic receiver coil determining the position of the sensor in the electromagnetical field is placed at the needle’s tip (black arrow). A cable connects the mandrin’s rear end with the EMT system.

Figure 6. Screenshots of the navigation system during planning (a), beginning (b), and end (c) of puncture. During planning different axial and sagittal reformations (a, upper half) allow for target identification while reformations along the planned trajectory allow for identification of anatomical structures along the puncture tract (a, lower half). During puncture the system informs visually (b, c) via concentric circles/markers in the image and quantitatively (b, c, numbers lower left) about the needle tip position and deviation from the planning. As needle is advanced the marquee-like display informs about the distance to the target (arrows in b, c).

Having reached the target position (Fig. 6), the puncture track was marked by using pieces of wire (Fig. 4). Planning and puncture times as well as puncture lengths were recorded. After the punctures RMFP imaging was performed for documentation and evaluation purposes. The deviations from the anatomical targets were evaluated by using multiplanar reconstructions. All procedures were carried out three times; planning and puncture times were recorded for the first pass.

Ex vivo animal model
Two domestic pig spine and back cadaver parts were placed on the RMFP patient support with the reference plate underneath the animal. Analogous to the phantom experiments lumbar facet joints and intervertebral discs were punctured by using electromagnetical tracking based on RMFP planning. Distances from the anatomical target were assessed by using RMFP control images. Punctures were repeated three times for each cadaver; for each first pass planning and puncture times were recorded.

Statistical analysis
Statistical analyses were performed by using PASW 17 (SPSS Inc., Chicago, Illinois, USA) employing non- parametric testing (Mann–Whitney U test). A p value of less than 0.05 was considered statistically significant.

Results
A total of n=210 RMFP imaging based EMT tracked punctures were carried out, n=90 in the phantom setting, 120 in the ex vivo spine model. Planning and puncture time data were recorded for 20 phantom based punctures and 40 ex vivo spine punctures.

For facet joint punctures the ex vivo model punctures were slightly more accurate (0.8±0.9 mm vs. 0.6 ± 0.8 mm) at a shorter puncture distance (38.8 ± 1.9 mm vs. 46.9 ± 3.9 mm, p < 0.001) although this difference was not statistically significant (p = 0.581). While planning times were almost identical (phantom, 19.7 ± 5.1 s; ex vivo, 18.6 ± 5.0 s) at no statistical difference (p=0.627) puncture times were significantly longer for ex vivo procedures (37.9±9.0 s vs. 23.6± 7.2 s, p=0.001).

Similarly intervertebral disc punctures were comparably accurate (phantom, 0.6±1.2 mm; ex vivo, 0.5±0.6 mm) with no significant difference (p = 0.142) at significantly longer puncture distances for the phantom based procedures (84.9± 6.2 mm vs. 67.4 ± 3.3 mm, p < 0.001). For both puncture and planning times a significant difference in favour of the phantom based punctures could be found (planning, 20.8± 5.0 s and 39.3 ± 17.3 s with p = 0.001; puncture, 31.1 ± 6.4 s vs. 43.9±16.1 s with p=0.026).

Table 1. Summarised data for the performed punctures. Dist = puncture distance, Dev = deviation from target, T (Pl) = planning time, T (Pu) = puncture time, SD = standard deviation.

Comparing direct with angular punctures for phantom model based facet joint punctures no difference in puncture accuracy (direct, 0.8 ± 0.9 mm; angulated, 1.0 ± 1.3 mm, p= 0.780) was found; although angulated puncture accuracy was slightly poorer, puncture distances were significantly longer for angulated punctures (46.9±3.9 mm vs. 57.3± 5.0 mm, p < 0.001). While puncture times were similar (direct, 23.6 ± 7.2 s; angulated, 19.6 ± 5.0 s, p = 0.102) planning times were shorter for the direct approach (19.7± 5.1 s vs. 28.6±7.8 s). Data are summarised in Table 1 and statistics are included in Table 2.

Table 2: Statistical comparisons of assessed data. Compared are phantom and ex vivo procedures as well as angulated versus direct punctures with respect to puncture distance, deviation from target, planning and puncture times. For all tests non-parametric Mann-Whitney U testing was performed. Statistically significant values are marked with an asterisk (*); other values are non-significant (p=0.05). The sign of the difference is given in parentheses. Dist = puncture distance, Dev = deviation from target, T (Pl) = planning time, T (Pu) = puncture time, FJ = facet joint puncture, IVD = intervertebral disc puncture, ph = phantom, ex = ex vivo, an = angulated, di = direct.

Discussion
Electromagnetic tracking provides an elegant and accurate way to control the position of an instrument’s tip based on imaging data^4,8,9. However, while CT based imaging provides the large data volumes with high spatial and temporal resolutioⁿ⁴, C-arm based CT approaches seem advantageous for EMT based interventions as a metal free environment during interventions rules out any magnetic field disturbances caused by e.g. the CT gantry or the CT table driving mechanism¹⁰.

Most C-arm CT systems, however, had the drawback of slow imaging times with limited anatomical coverage, negatively influencing suitability of such systems for navigated procedures as imaging needs to cover the whole region of the lesion and adjacent tissues and a referencing system is necessary in most cases.

RMFP systems provide ideal conditions for EMT based procedures as (a) large imaging volumes (40 _ 40 _ 24 cm) can be achieved in CT grade quality with reasonable data acquisition times (˜ 15 s); (b) any metallic parts can be swiftly moved out of the interventional volume without displacing the patient; (c) free patient access is available when the RMFP system is parked; (d) fallback projectional imaging is provided by with a large high-performance flat panel detector (30 _ 40 cm). While RMFP can provide higher spatial resolution in comparison to standard CT, the drawbacks are the slower imaging, especially for contrast-enhanced imaging, and the higher radiation exposure in comparison to volume acquisitions performed by standard C-arm CT systems⁶.

A system featuring similar detector capabilities and specifications is the fully covered O-arm system (Medtronic, Boulder, Colorado, USA)11. While having the advantage of lacking moving parts near the patient or interventionalist it features only a limited parking capability in comparison to the system used in the study. While similar performance for EMT based procedures can be expected a thorough systematic comparison is still outstanding.

The achieved accuracies of 0.7±0.9 mm (0.5-1.0±0.6-1.3 mm) are comparable to the accuracies measured by Bruners et al.⁴ with 0.4±0.8 mm in an EMT-guided CT- based phantom setting, in the range of the technical error of the system (0.6 mm)¹², and more accurate than the 2.8± 2.1 mm assessed in an ex vivo model by the same group. Possible reasons are more reliable fixture of the registration plate in the ex vivo setting when using a carbon based table, where it can be placed underneath the patient and no fixture to the skin is necessary.

In vivo tests and clinical use of EMT systems provided puncture accuracies of 3.0±2.0 mm (in vivo pig⁴) and 5.4±1.9 mm (clinical setting⁸), the most probable reason for that discrepancy being the effect of respiratory motion, which even under breath hold could lead to a slight registration error. In a standard C-arm based phantom setting accuracies of 2.3±0.9 mm could be shown⁸, which are inferior to the results of this study, although here the comparability of the phantoms has to be questioned.

No differences in accuracy between phantom and ex vivo or angulated and direct punctures could be detected at reasonably high puncture numbers (n = 30 and n = 60) while significant differences in the puncture lengths could be detected for the different models and puncture types, suggesting a minor influence of puncture length on procedure accuracy. This is probably owing to the constantly monitored instrument tip eliminating needle bending as a cause of error.

Puncture times were longer for the ex vivo based procedures probably caused by the higher mechanical resistance caused by tissue in comparison to gel wax. Planning times were significantly longer for ex vivo based intervertebral disc punctures when compared to the phantom interventions, maybe caused by the slightly different anatomy of the ex vivo spine in comparison to the anthropomorphic model with narrower intervertebral disc spaces. The longer planning for angulated phantom based facet joint punctures is probably explained by the more complex planning procedure for angulated punctures.

Conclusion
The combination of a robot mounted flat panel system with an electromagnetic tracking system provides an accurate method of navigation for spinal interventions such as facet joint punctures and intervertebral disc punctures. As a phantom and ex vivo setting was performed, further clinical studies will provide further insight into the routine applicability of the system combination.

Acknowledgements
:
This work has been funded in part by the German Ministry for Education and Research (BMBF) in the framework of the OrthoMIT project under grant no. 01EQ0402. Equipment and material for this study were provided by CAS innovations, Siemens Healthcare, Erlangen, Germany. The authors collaborate in research with Medtronic.

References:

1. Bale R, Widmann G (2007) Navigated CT-guided interventions. Minim Invasive Ther Allied Technol 16:196– 204. doi:10.1080/13645700701520578
2. Silverman SG, Tuncali K, Adams DF, Nawfel RD, Zou KH, Judy PF (1999) CT fluoroscopy-guided abdominal inter- ventions: techniques, results, and radiation exposure. Radiology 212:673–681
3. Krombach GA, Mahnken A, Tacke J, Staatz G, Haller S, Nolte-Ernst ing CC, Meyer J, Haage P, Gunther RW (2001) US-guided nephrostomy with the aid of a magnetic field-based navigation device in the porcine pelvicaliceal system. J Vasc Interv Radiol 12:623– 628
4. Bruners P, Penzkofer T, Nagel M, Elfring R, Gronloh N, Schmitz- Rode T, Günther RW, Mahnken AH (2009) Electromagnetic tracking for CT-guided spine interventions: phantom, ex-vivo and in-vivo results. Eur Radiol 19:990– 994. doi:10.1007/s00330-008-1227-z
5. Nagel M, Hoheisel M, Bill U, Klingenbeck-Regn K, Kalender WA, Petzold R (2008) Electromagnetic tracking system for minimal invasive interventions using a C-arm system with CT option: first clinical results. In: Medical imaging 2008: visualization, image-guided procedures, and modeling. SPIE, San Diego
6. Kalender WA, Kyriakou Y (2007) Flat- detector computed tomography (FD- CT). Eur Radiol 17:2767–2779. doi:10.1007/s00330-007-0651-9
7. Vogl TJ, Naguib NN, Nour-Eldin NE, Lehnert T, Mbalisike E (2009) C-arm computed tomography for transarterial chemoperfusion and che moembolization of thoracic lesions. Radiologe 49:837–841. doi:10.1007/s00117-009-1863-6
8. Meyer BC, Peter O, Nagel M, Hoheisel M, Frericks BB, Wolf KJ, Wacker FK (2008) Electromagnetic field-based navigation for percutaneous punctures on C-arm CT: experimental evaluation and clinical application. Eur Radiol 18:2855–2864. doi:10.1007/s00330- 008-1063-1
9. Wood BJ, Zhang H, Durrani A, Glossop N, Ranjan S, Lindisch D, Levy E, Banovac F, Borgert J, Krueger S, Kruecker J, Viswanathan A, Cleary K (2005) Navigation with electromagnetic tracking for interventional radiology procedures: a feasibility study. J Vasc Interv Radiol 16:493–505. doi:10.1097/ 01.RVI.0000148827.62296.B4
10. Yaniv Z, Wilson E, Lindisch D, Cleary K (2009) Electromagnetic track ing in the clinical environment. Med Phys 36:876–892
11. Zhang J, Weir V, Fajardo L, Lin J, Hsiung H, Ritenour ER (2009) Dosi metric characterization of a cone- beam O-arm imaging system. J Xray Sci Technol 17:305–317. doi:10.3233/XST- 2009-0231
12. Nagel M, Hoheisel M, Petzold R, Kalender WA, Krause UHW (2007) Needle and catheter navigation using electromagnetic tracking for computer-assisted C-arm CT interventions. In: Medical imaging: visualization and image-guided procedures. SPIE, San Diego.