1 Introduction

The O-arm Surgical Imaging System (Medtronic, Louisville, CO, USA) is a multidimensional surgical imaging platform designed for use in surgeries of the spine and brain, and in cases of trauma. The system provides real-time intraoperative imaging of the patient's anatomy through high-quality images and two-/three-dimensional (2D/3D) fields-of-view. By integrating the O-arm surgical imaging technique and the Stealth Station Surgical Navigation System (Medtronic), a surgeon can improve the visibility of complex procedures and ensure surgical accuracy. Our hospital introduced the O-arm and Stealth Station systems in November 2014 and 60 spinal operations have been performed using these instruments to assess their safety and precision. This report describes our experience.

2 Methods and materials 2.1 General information

The study enrolled 60 spinal internal fixation patients (36 males, 24 females; age range, 8-76 years) who were treated and followed-up between May 1, 2015 and July 1, 2017. The treated conditions included atlantoaxial subluxation (28 cases), cervical kyphosis (4), thoracic intraspinal meningioma (4), lumbar spinal stenosis (8), lumbar spondylolisthesis (10), lumbar vertebral compression fracture (4), lumbar spinal schwannoma (1), and sacrococcygeal chordoma (1). The patients were classified into two groups, according to their lesion type: the occipitocervical fusion and posterior neck fixation group and the thoracic/lumbar/sacral pedicle internal fixation group.

2.2 Methods 2.2.1 Imaging examination

Before surgery, all patients underwent plain radiography, computed tomography (CT), and plain and enhanced-scan magnetic resonance imaging (MRI).The lesions were well documented. Intraoperatively, 3D O-arm anteroposterior and lateral views were obtained and recorded. Either radiography or CT was repeated 1 month postoperatively to confirm implant locations.

2.2.2 Navigation method

The O-arm intraoperative imaging system and Stealth Station^® S7 surgical navigation system were used. The former included an imaging workstation and ray transmitter, and the later included a computer workstation, navigation system, and infrared positioning system (including reference rack, navigation tracker, wireless navigation guide bar, and adapter). The working principle involves the reference rack being placed on the spinous process of an upper or lower segment of a fixed vertebral body to ensure relative location fixation of the spinous process and reference rack. This also ensures that the infrared positioning signals and the reference rack's receiving device are optimally aligned. After O-arm scanning, the 3D images are imported into the navigation system. After the wireless navigation guide bar has successfully registered in the reference rack registry, the navigation guide bar accuracy is corrected at each exact spinous process locus. After accuracy is established, intuitive real-time navigation is performed.

2.2.3 Surgical method

The surgery was performed under general anesthesia.The operating table was a Jackson multi-purpose perspective spine bed. The navigation system was placed at the head of the operating table, and the workstation was placed on the side opposite the surgeon.A technician connected the O-arm and navigation systems and entered the patient's information into the system. Before the operation, the O-arm's 2D perspective mode was used for positioning.

Each patient in the occipitocervical fusion group was placed prone, with the head under fixed traction. The occipital bone and cervical posterior structure were visualized through the conventional posteromedial approach. First, posterior cranial fossa suboccipital decompression was performed microscopically, and the navigation reference rack was fixed to the lower vertebral spinous process root of the plant screw. The occipital bone plate was fixed with occipital screws, C2 was fixed with Pars, and the C3, C4, and/or C5 vertebrae were fixed with lateral blocks.

Each patient in the thoracic/lumbar/sacral pedicle internal fixation group was also placed prone. The thoracic/lumbar/sacral posterior structures were revealed through a conventional posteromedial approach, and the navigation reference rack was fixed to the upper vertebral spinous process root of the plant screw. Bilateral pedicle fixation was selected, according to the disease. O-arm scanning produced 2D and 3D anteroposterior and lateral images, which were transmitted to the navigation system workstation. The accuracy of the navigation was verified on the spinous process after the navigation guide bar was registered on the reference rack. The navigator provided many implantation orbits, and the surgeon selected the best one, according to the navigation guide. Screw diameters, lengths, and paths were selected under navigational guidance, and each screw was implanted using real-time navigation. After screw implantation, O-arm anteroposterior and lateral 2D and 3D images were again obtained to evaluate the positioning, direction, and depth of each implant.

2.2.4 Evaluation method

Screw position accuracy was determined with reference to Richter's four classification criteria. (1) We evaluated the accuracy of the lateral block or pedicle screws according to the patient's intraoperative O-arm 3D images, assigning one of three grades. Grade Ⅰ: the screw position is satisfactory—i.e., the screw did not penetrate the lateral block or pedicle cortex, or it penetrated one or the other only slightly ( < 1 mm).Grade Ⅱ: the screw penetrated the lateral block or pedicle cortex by >1 mm, but did not need to be adjusted. Also, the patient showed no peripheral tissue injury and the screw was stable. Grade Ⅲ: the screw penetrated the lateral block or pedicle cortex by >1 mm, and there was peripheral tissue injury.Alternatively, the screw was unstable and needed to be adjusted. (2) Operation time: the average perscrew implantation time. (3) Frequency, duration, and navigation time of the O-arm scan. (4) The presence of complications, with the reasons for their occurrence.

3 Results

There were 32 patients (12 males, 20 females) requiring occipitocervical fusion and posterior neck fixation; a total of 182 cervical screws and 84 occipital screws were implanted. There were 28 patients (15 males, 13 females) requiring thoracic/lumbar/sacral pedicle internal fixation; 148 pedicle screws were implanted.

3.1 Pedicle screw position accuracy

In the occipitocervical fusion group, a total of 182 cervical screws were implanted with navigation assistance, including 64 C2 Pars screws and 118 C3 and/or C4 lateral block screws. Intraoperative O-arm 3D scans revealed 176 Richter grade Ⅰ screws and 6 grade Ⅰ screws penetrating the lateral block cortex near the vertebral artery (Figure 1), yielding an accuracy of 96.7% (176/182). Among the 84 occipital screws, 3 deviated from the occipital posterior midline position, yielding an accuracy of 96.4% (81/84).

For the thoracic/lumbar/sacral pedicle internal fixation group, 148 pedicle screws were implanted under navigational guidance. Intraoperative O-arm 2D lateral scans did not show screw position deviations, but the 3D scans found 148 Richter grade Ⅰ screws (0 screws penetrating the pedicle) and 4 grade Ⅱ screws that had deviated to the lateral pedicle (Figures 2-5).Intraoperative O-arm 3D scans were repeated for confirmation. Secondary correction was performed under navigational assistance, yielding a final accuracy of 97.3% (144/148).

3.2 Time and frequency of screw implantation

In the occipitocervical fusion group, the navigationguided implantation operation time for each cervical spine screw was 5-7 min, with a total implantation time for each occipital screw of 3-5 min. In the thoracic/lumbar/sacral pedicle fixation group, the navigationguided implantation operation time for each thoracic pedicle screw was 8-10 min, with an implantation time for each lumbar pedicle screw of 6-8 min.

3.3 O-arm scan frequency, duration, and navigational time

The preparation time for starting the O-arm and its navigational equipment, setting parameters, and connecting was 20-30 min/time. The average time for an O-arm anteroposterior and lateral 3D scan and importing these into the navigational system was 3-5 min/time. The average O-arm scan frequency was 3 times/case (including positioning the section, 3D scanning, importing the image into the navigation system, and post-implantation verification). The radiation dose for each 3D scan was approximately 5-7 mSv.

3.4 Immediate complications

Postoperatively, one occipitocervical fusion patient developed dysphagia and underwent removal of the connecting rod after no improvement was observed within 2 months. The deglutition function gradually recovered over 1 month. This complication is believed to be related to over-flexion of the neck after occipitocervical fusion.

4 Discussion

Boucher et al. first applied the spinal pedicle fusion technique in 1959. Since then, the technique has become prevalent worldwide. However, empirical "blind screwing" positional deviations of approximately 10%-40%^[1] result in damage to the nerve/spinal cord/ blood vessels and even to intra-abdominal organs, making the accuracy of pedicle screw implantation a challenge for clinicians. Roberts et al. first applied a neurosurgical navigation technique in 1986^[2], which was extended to spinal fixation by Steinmann et al.^[3], greatly improving screw implantation accuracy. The current navigation techniques include preoperative CT navigation, isocenter C-arm navigation, and intraoperative O-arm navigation. The latter technique, intraoperative O-arm real-time navigation, is the most advanced of these techniques; however, reaching 100% accuracy is difficult, regardless of the navigation technique employed. We studied the 60 intraoperative O-arm real-time navigation cases performed in our department and analyzed the possible causes of navigation inaccuracy and screw implantation bias.

Much of the literature states that the accuracy of navigation-assisted screw fixation is >90%, which is higher than that of traditional screw implantation, improving safety. Cong et al. examined nine C3-7 pedicle screws implantations involving navigational assistance, and found an accuracy of 90%, compared with an accuracy of 55.6% for non-assisted implantations^[4]. Yang et al. reported that the accuracy of C1-2 articular screw fixation was 97.2% in 36 cases involving isocenter C-arm navigation, compared with an accuracy of 91.7% for 48 conventional radiography cases^[5]. Tao et al. reported that 99 patients underwent intraoperative three-dimensional fluoroscopic navigation, with an accuracy of 98.5% for 196 C3 screw implantations^[6]. In this group of occipitocervical fusion patients, the accuracy of performing cervical lateral blocks and implanting Pars screws was 96.7%. Screws penetrated the bone cortex in six patients, resulting in the screws being near, but not injuring, the vertebral arteries; postoperative neurological dysfunction was not observed. These results are similar to the literature results already discussed.

Screw penetration may be related to excessive screw lengths and to the implantation angle. Although depth measurements may be performed under navigation, "individualizing" the screw model is difficult. Screws that are too short may not achieve biological stability, and those that are too long may result in penetrations. In the future, custom 3D-printed screws may solve this problem. In addition, the deviation between the actual spinal position and the navigational image also leads to decreased accuracy, especially in patients requiring multi-segment screw implantations. Hence, when the intraoperative experience differs from expectations, repeated identification and calibration of the navigation system accuracy is necessary.

Laine et al. reported the accuracy of thoracolumbar and lumbosacral pedicle screws in 100 cases and noted that the ratio between the navigational implantation group and the traditional implantation group was 95.4%:86.6%^[7]. Amiot et al. reported a total of 294 thoracic/lumbar/sacral screws in 50 patients in a navigation implantation group and 544 thoracic/ lumbar/sacral screws in 100 patients in a conventional implantation group. Their accuracy rates were 95% and 84%, respectively^[8]. Bledsoe et al. reported 150 thoracic pedicle screw implantations in 34 patients, using isocenter C-arm navigation, with an accuracy of 93.3%^[9]. The implantation of 278 thoracic/lumbar/ sacral pedicle screws, under intraoperative O-arm navigation, was previously reported to have an accuracy of 96.8%^[10]. Further, the accuracy of implanting 68 lumbar pedicle screws, under intraoperative O-arm navigation, was previously reported to be 91.2%^[11].

Kosmopoulos et al. retrieved 130 articles and performed a meta-analysis of the accuracy of implanting 37, 337 pedicle screws. Their results showed that the accuracy of screw implantations in the navigation group was 95.2%, and that in the non-navigation group was 90.3%^[12]. For the thoracic/lumbar/sacral vertebrae fusion patients in our group, the accuracy of pedicle screw implantation was 97.3%; four screws moved out of the pedicle but did not cause neurological dysfunction. Our results are very similar to those reported in the discussed articles. Therefore, based on the results of this group and previous literature reports, intraoperative screw implantation, under O-arm real-time navigation, is highly accurate and has correspondingly good safety.

The accuracy of screw implantations involving O-arm navigation did not reach 100%, and the screw skewing may be related to several factors, including (1) the difficulty of positioning screws in the vertebral body so that they are completely consistent with the position scanned by the O-arm, considering the impacts of breathing and implantation force; (2) developmental deformities and variations in the lateral blocks or pedicles, vertebral instability, screw slippage; and (3) surgical equipment operation, screw implantation, and navigation system errors.

From the perspective of screw implementation and equipment operation times for the two groups, the average O-arm scan frequency was 3 times/case, with a total duration of equipment use of approximately 25-35 min and an average screw implantation time of 3-10 min. Despite the absence of a traditional screw implementation control group, the frequency of and time required for screw position verification using traditional repeated fluoroscopic guidance would normally be longer. In addition, radiation exposure during screw implantation is an unavoidable clinical problem for both patients and medical staff. Although experienced surgeons can significantly reduce the frequency and duration of fluoroscopic examinations, the frequency and total radiation dose are significantly greater for new doctors or patients suffering from spinal deformities. Lange et al. reported that the radiation dose delivered during an O-arm 3D scan of the thoracic and lumbar spine is about 3-8 mSv. In contrast, the radiation dose associated with a spiral CT scan of the thoracic and lumbar spine is about 18-19 mSv^[13]. Zhang et al. reported that the radiation dose of the O-arm in 3D mode is only half that associated with 64-row CT^[14]. In this study, the average radiation dose associated with an O-arm 3D scan was approximately 5-7 mSv. Thus, the reduced dose and frequency of imaging greatly reduced the radiation exposure for both patients and medical staff. For long-segment screw implantation surgeries, however, the radiation exposure increases by two or more times that typically associated with 3D scanning and the transmission of data into the navigation system also increases due to O-arm scanning range restrictions. Together, these factors increase operation duration and radiation exposure, indicating the need for further technical refinements.

In short, screw implantation involving O-arm realtime navigation allows observation of the screw angle and depth in real-time, greatly improving implantation accuracy and safety, reducing surgeon and patient exposure to radiation, and shortening the time required for new doctors to learn screw implantation techniques. Finally, these types of navigation techniques will be useful for percutaneous, minimally invasive screw implantations and for endoscopic techniques, which are gradually increasing in frequency. This technique further improves the accuracy of a minimally invasive technology.

Acknowledgements

This project was supported by the Beijing Tsinghua Changgung Hospital Fund (No. 12015C1045). The authors thank Peihai Zhang and Linkai Jing for collecting the archived data and Guihuai Wang and Jin Wang for technical help.

Conflict of interests

None of the contributing authors have conflicts of interest.