Key points

History of image-guided instrumentation of the spine

Early instrumented spinal fusion consisted of rod constructs with sublaminar wiring or hooks, which were placed with nearly direct visualization. Modern three-column fixation with pedicle screws increases fusion rates and provides greater stability, but involves placing instrumentation into unexposed anatomy without direct visualization of the trajectory. Improper placement can lead to complications, such as nerve root impingement, spinal cord compromise, great vessel injury, or an unstable construct that may necessitate further surgery. Two-dimensional (2D) fluoroscopy with a C-arm is the traditional adjuvant to a thorough understanding of the anatomic relationships in the spine, but this is limited by several factors: (1) the surgeon can examine only two planes, anteroposterior (AP) and lateral, and only one at a time; (2) the need for multiple repeated fluoroscopic images in different planes and at each level increases the radiation exposure to the surgeon, operating room (OR) staff, and patient; (3) the size and awkward shape of the C-arm can constrict the surgeon’s and assistant’s movement; and (4) the image quality is poor in the lower cervical and upper thoracic spine.

The limitation of 2D fluoroscopy was the driving factor for the invention of computed tomography (CT), and the Nobel Prize in medicine and physiology was awarded to Hounsfield and Cormack in 1979 for their independent but nearly simultaneous inventions of this technology. Preoperative CT scans gave the spine surgeon superior insight into the patient’s individual spinal anatomy, allowing more thorough planning that involved predetermining appropriate implant size and length, and consideration of any pathologic conditions that might have distorted the normal anatomy. To minimize radiation from repeated 2D fluoroscopic images, early incorporation of CT-guided spine surgery used preoperative CT with passive navigation during surgery. In one study, this was found to decrease erroneous pedicle screw placement from 44% by a freehand technique to 9% with the use of preoperative imaging. In the early 1990s, a mobile CT scanner became available, and with it the opportunity to bring three-dimensional (3D) imaging into the OR. However, early mobile scanners were cumbersome and had little intraoperative utility until a decade later when the O-arm (Medtronic, Memphis, TN) and Iso-Centric (Siemens, Malvern, PA) systems were introduced. The open gantry concept afforded greater maneuverability and access to the OR table for instant intraoperative image acquisition. Although 3D navigation may initially extend operative time because of the registration and image processing time, and the associated learning curve, one surgeon’s series of one-level fusion in 133 patients, 63 with freehand pedicle screw placement and 70 with 3D navigation–assisted placement, showed that the operative time for navigated surgeries was an average of 23 minutes shorter.

Intraoperative 3D neuronavigation is now widely available, and proponents claim this can increase instrumentation accuracy, decrease radiation exposure, shorten OR time, and improve patient outcomes. However, as with any technical innovation, there is a large capital expenditure necessary to implement these systems. Before making this investment, the surgeon should have a thorough understanding of the advantages and limitations of 3D image–guided instrumentation; and acknowledge that 3D navigation is an adjuvant to, rather than a replacement of, anatomic knowledge. With these proposed advantages in mind, this article critically appraises the available literature to assess the influence of 3D navigation on radiation exposure, accuracy of instrumentation, operative time, and patient outcomes. We also explore the latest technological advance in 3D neuronavigation: the manufacturing of, via 3D printers, patient-specific templates that direct implant placement.

History of image-guided instrumentation of the spine

Early instrumented spinal fusion consisted of rod constructs with sublaminar wiring or hooks, which were placed with nearly direct visualization. Modern three-column fixation with pedicle screws increases fusion rates and provides greater stability, but involves placing instrumentation into unexposed anatomy without direct visualization of the trajectory. Improper placement can lead to complications, such as nerve root impingement, spinal cord compromise, great vessel injury, or an unstable construct that may necessitate further surgery. Two-dimensional (2D) fluoroscopy with a C-arm is the traditional adjuvant to a thorough understanding of the anatomic relationships in the spine, but this is limited by several factors: (1) the surgeon can examine only two planes, anteroposterior (AP) and lateral, and only one at a time; (2) the need for multiple repeated fluoroscopic images in different planes and at each level increases the radiation exposure to the surgeon, operating room (OR) staff, and patient; (3) the size and awkward shape of the C-arm can constrict the surgeon’s and assistant’s movement; and (4) the image quality is poor in the lower cervical and upper thoracic spine.

The limitation of 2D fluoroscopy was the driving factor for the invention of computed tomography (CT), and the Nobel Prize in medicine and physiology was awarded to Hounsfield and Cormack in 1979 for their independent but nearly simultaneous inventions of this technology. Preoperative CT scans gave the spine surgeon superior insight into the patient’s individual spinal anatomy, allowing more thorough planning that involved predetermining appropriate implant size and length, and consideration of any pathologic conditions that might have distorted the normal anatomy. To minimize radiation from repeated 2D fluoroscopic images, early incorporation of CT-guided spine surgery used preoperative CT with passive navigation during surgery. In one study, this was found to decrease erroneous pedicle screw placement from 44% by a freehand technique to 9% with the use of preoperative imaging. In the early 1990s, a mobile CT scanner became available, and with it the opportunity to bring three-dimensional (3D) imaging into the OR. However, early mobile scanners were cumbersome and had little intraoperative utility until a decade later when the O-arm (Medtronic, Memphis, TN) and Iso-Centric (Siemens, Malvern, PA) systems were introduced. The open gantry concept afforded greater maneuverability and access to the OR table for instant intraoperative image acquisition. Although 3D navigation may initially extend operative time because of the registration and image processing time, and the associated learning curve, one surgeon’s series of one-level fusion in 133 patients, 63 with freehand pedicle screw placement and 70 with 3D navigation–assisted placement, showed that the operative time for navigated surgeries was an average of 23 minutes shorter.

Intraoperative 3D neuronavigation is now widely available, and proponents claim this can increase instrumentation accuracy, decrease radiation exposure, shorten OR time, and improve patient outcomes. However, as with any technical innovation, there is a large capital expenditure necessary to implement these systems. Before making this investment, the surgeon should have a thorough understanding of the advantages and limitations of 3D image–guided instrumentation; and acknowledge that 3D navigation is an adjuvant to, rather than a replacement of, anatomic knowledge. With these proposed advantages in mind, this article critically appraises the available literature to assess the influence of 3D navigation on radiation exposure, accuracy of instrumentation, operative time, and patient outcomes. We also explore the latest technological advance in 3D neuronavigation: the manufacturing of, via 3D printers, patient-specific templates that direct implant placement.

Three-dimensional navigation in adult spine surgery: techniques

The components of navigation include the patient image, obtained either preoperatively or intraoperatively; the computer workstation with appropriate software to process the image and create 3D reconstructions; an optical camera; instrument trackers; and the reference point, which is placed on a rigid anatomic structure, such as an adjacent spinous process or the posterior superior iliac spine in percutaneous cases ( Fig. 1 ). After surgical exposure of the appropriate anatomy, we cover the operative field with two sterile drapes secured together with towel clamps, and then cover the reference frame with a clear sterile drape. The open gantry CT scanner is brought into the field and closed with the patient in the isocenter under direct visualization by the surgical team to avoid contact with the field. The technician operating the intraoperative CT secures these additional sterile drapes underneath the operating table; the drapes tightly fit the patient and cannot get caught or displaced as the scanner moves. AP and lateral radiographs confirm the levels of interest are included in the spin. Members of the surgical team step into a substerile zone, the anesthesiologist provides apnea, and the technician initiates the spin, which concludes in 40 seconds for high-definition at our institution. Following the spin, one member of the surgical team removes the outer drapes and changes their outer gloves before registering the navigated instruments. If desired, the intraoperative CT is fused to a preoperative scan that projects preplanned screw sizes and trajectories. Otherwise, navigation is performed using the intraoperative scan alone, as is the case in our institution. The camera, positioned either at the head or foot of the bed, emits infrared light and detects the light from reflecting spheres on the reference frame and on each navigated instrument. The software determines the multiplanar projections of each instrument and projects this onto a workstation for the surgeon to view during instrumentation placement. To confirm accuracy, the surgeon navigates on known anatomic landmarks before placing instrumentation.

( A ) Intraoperative set up for a spinal procedure with the patient positioned prone, draped, and two computer workstation monitors at the head of the operating table in view of both surgeons. ( B ) A sterilely draped reference frame affixed to a lumbar spinous process. ( C ) The optical camera is positioned behind the anesthesia drapes with view of the exposed surgical field. ( D ) The patient is covered with an additional temporary sterile drape, and an open gantry intraoperative CT scanner is brought into the operative field. The anesthesiologist suspends respirations and a 40-second image acquisition spin is obtained for a high-resolution scan. ( E ) Example of a high-resolution intraoperative CT scan acquired during a spinal deformity surgery. ( F ) A navigated probe with reference markers is registered with the reference frame affixed to the spinous process. ( G ) The navigated probe is placed on various exposed anatomic landmarks (eg, spinous process, transverse process) and confirmed with the navigated image displayed on the computer screen. ( H ) Conventional surgical instruments (eg, pedicle screw probes, taps, drivers) are affixed with reference markers for navigated assistance of each step of placing spinal instrumentation. ( I ) Projected trajectory of a lumbar pedicle screw is seen on the computer workstation during creation of the pedicle screw tract, tapping, and screw insertion. After all instrumentation is inserted, a final intraoperative CT scan is performed to confirm accurate placement.

Three-dimensional navigation in adult spine surgery: outcomes

Now that 3D navigation has been common practice for many surgeons, patient series are being published about the accuracy of implant placement. Verma and colleagues did a retrospective analysis of 537 patients from a high-volume trauma center that had suffered lumbar spine fractures requiring instrumented stabilization. Spine trauma poses an additional challenge for the surgeon, because the normal anatomy is often distorted. In 278 patients, screw placement was performed using 3D navigation, and in 309 patients, pedicle screws were placed using 2D C-arm navigation. The authors found a statistically significant difference, with approximately 1% of screws being misplaced in the 3D navigation group compared with approximately 9% in the 2D navigation group. Their data suggest that 3D image guidance is especially useful in the setting of trauma, when instability may cause anatomic relationships to change from when the patient is positioned supine in the CT scanner compared with prone on the operating table.

In a prospective, randomized, observational study, Noriega and colleagues found similar results in thoracolumbar degenerative spinal fusion. They included 114 consecutive patients in their study, 56 underwent instrumented fusion using free-hand fluoroscopy-guidance and 58 patients were treated using 3D navigation. Perforation rates, as measured on postoperative CT scans by the Heary classification, were significantly higher in the free-hand group (10.3%) compared with the CT-navigated group (3.6%). The authors also investigated radiation exposure, using thermoluminescent dosimeters to quantify cumulative skin dose and using data from each imaging apparatus to calculate the effective dose and organ dose to each patient. The group of patients undergoing 3D navigated instrumentation had a higher radiation dose, but the surgical team had a lower radiation dose. In a risk versus benefit analysis, the authors argue that the absolute reduction in pedicle perforation may warrant the increased radiation exposure to the patient, because the effective radiation dose was still within known safety limits for the risk of carcinogenesis. However, note that all breaches were asymptomatic in this study except for one in the 2D navigated group. Long-term follow-up studies are needed for the surgical community to establish the clinical relevance, if any, of an asymptomatic pedicle breach.

Another group, Uehara and colleagues, recently published their series of 359 consecutive patients who underwent pedicle screw fixation using 3D navigation. This patient population was more heterogeneous than the one described previously, because it also included patients undergoing cervicothoracic fusion. The authors obtained postoperative CT scans and quantified perforation rates. Of the 3413 pedicle screws inserted, approximately 7% were found to have perforated the pedicle. The authors found that instrumenting the mid-cervical and upper and mid-thoracic spine was associated with higher perforation rates (9%–11%) when compared with the lumbar spine (3.8%). These higher perforation rates highlight that 3D navigation can be misleading in areas that are anatomically challenging, and that it should not be counted on as the sole determinant of safe implant placement.

As different approaches to spinal fusion evolve, so will the applications for 3D navigation. In addition to pedicle screw fixation, current spinal fusion procedures often involve an interbody cage. These implants restore disk height, open the neural foramen to relieve radiculopathy, and facilitate bony fusion, but preparation of the disk space and delivery of the implant puts the thecal sac and nerve roots at risk. A misplaced interbody graft may compromise the canal or result in failure of the construct or of deformity correction. Some groups are now using 3D image guidance to ensure proper placement. Joseph and colleagues used CT-guided navigation for a minimally invasive lateral approach to lumbar interbody fusion and found that 64 of 66 cages were properly placed using 3D navigation. Intraoperative CT-based spinal navigation is also being explored in other challenging neurosurgical procedures, including odontoid screw placement.

Atlantoaxial fixation is uniquely complex because of the proximity of the vertebral arteries. Early surgical stabilization included dorsal wiring techniques, in which hardware was placed with direct visualization of the implant’s course. Dorsal wiring techniques, including Gallie, Brooks and Jenkins, and Sonntag’s modified Gallie fusion provide translational stability and are easy to perform. However, these fusion techniques require postoperative halo immobilization and are limited by (1) the need for an intact C1 posterior arch, (2) lack of rotational stability, and (3) suboptimal fusion rates. Three-column fixation, as in other areas of the spine, provides immediate stabilization, results in higher fusion rates, and adds rotational stability. With the addition of screw fixation techniques, dorsal wiring is now used mainly as an adjunct, rather than a primary means of stabilization. Magerl’s technique uses a single transarticular screw that enters 4-mm rostral to the C2-3 facet and ends at the anterior tubercle of C1. The screw courses adjacent to the vertebral foramen of C2, placing the vertebral artery at risk. To better define the preoperative safety of a transarticular screw, Paramore and coworkers used early 3D CT reconstructions to create a single image of the screw’s path. The authors reviewed scans from 94 patients, and found the anatomy unacceptable for screw placement in 17 patients (18%). With this knowledge preoperatively, the authors revised their surgical plan and placed a unilateral screw, augmented the fusion with dorsal wiring, and placed the patient in a halo postoperatively.

Traditionally, 2D fluoroscopy was used to evaluate patients before transarticular screw placement, and those with C2 pars width and/or height less than 5 mm were excluded. As technology progressed to intraoperative navigation based on 3D image acquisition, screw placement could be performed safely even in patients previously determined to be poor candidates because of a narrow corridor for the screw. Bloch and colleagues established that 3.5-mm transarticular screws could be safely placed if the C2 pars height and width was greater than 4 mm, if using 3D navigation. Recently, Yang and colleagues compared 3D navigation with 2D fluoroscopy and achieved a 97% accuracy rate compared with 92%, respectively. The authors also found the total radiation time was reduced in the 3D navigated group.

When anatomic considerations prevent safe transarticular screw placement, and the patient is not a candidate for unilateral screw placement with postoperative halo immobilization, the rod cantilever technique is a suitable alternative with similar outcomes. In this technique, rods are fixed to individual polyaxial C1 lateral mass screws and C2 pars screws. Literature directly comparing 2D versus 3D navigation for the Harms technique is limited, but seems to suggest improved accuracy, less fluoroscopy time, and less blood loss.