Introduction
Over the past 30 years, spine surgery has undergone a rapid transformation largely due to technological advances in image-guided navigation. In traditional open, freehand, instrumented surgery, the screw trajectory is estimated after exposing the screw entry point and the nearby relevant anatomical landmarks. This is to ensure that screws are correctly placed and injury to neurovascular structures is avoided. The freehand technique requires large incisions and causes significant tissue damage. In trauma or deformity surgery when the anatomy is altered, the normal trajectories can be distorted increasing the risk of screw misplacement. 3-D navigational techniques were introduced after the development of mobile fan-beam and cone-beam computed tomography (CT) devices and computers with fast processor speed and navigational software. 3-D real-time rendering simplifies the conceptualization of 3-D anatomy and offers a high degree of accuracy. The adaptation and evolution of image-guided navigation techniques have allowed a move toward minimally invasive surgery (MIS) where the lack of anatomical visualization can be compensated for with navigation [
1]. Several reports have shown an improved workflow and increases in the safety, accuracy, and efficiency of MIS procedures [
2‐
6]. However, the development of spinal navigation solutions has been slower than the cranial ones. The relatively complicated and time-consuming registration process of spinal navigation devices and the inherent risk for decreased accuracy with distance from the dynamic reference frame may provide part of the explanation [
4,
7,
8]. Despite efforts to implement novel registration and tracking methods [
9,
10], the traditional infrared, reference-based, outside-in, navigation devices still dominate the market. Commercially available navigation devices use an indirect method for spine tracking where optical hardware is used to recognize a dynamic reference frame with reflecting spheres in a predetermined geometrical pattern [
11‐
13]. Tracking is initiated by providing the navigation system with information on the position of the patient either through a user feedback procedure to match the patient’s anatomy to the preoperative images, or by using intraoperative automatic image registration [
14,
15]. The manual registration typically requires several minutes to complete the multi-step user feedback procedure necessary to achieve the required accuracy [
16‐
19]. Due to manual errors, or factors such as movement due to respiration, the registration procedure may need to be repeated before an acceptable accuracy is achieved. Repeating the registration during surgery may not be feasible. Thus, manual registration prolongs and complicates the surgical process, limiting surgical efficiency and the wider implementation of spinal navigation. These shortcomings can be improved by the automatic detection and registration of landmarks [
20].
However, automatic intraoperative image registration has only been possible when the imaging device and the navigation system share a common interface with integrated calibration [
14,
15].
The aim of this study was to validate a novel automatic registration solution designed to be independent of the 3D scanning device: Universal AIR (Brainlab AG, Munich, Germany), and compare its workflow and accuracy to a traditional Surface Matching registration method (Spine and Trauma Navigation and Registration SW, Brainlab AG, Munich, Germany).
Discussion
The accuracy data of the present study demonstrate non-inferiority of automatic image registration using Universal AIR with intraoperative CBCT imaging as compared to the Surface Matching technique. In addition, a post hoc test revealed an improved registration accuracy using Universal AIR. In clinical terms, the accuracies of the methods, 1.20 mm and 1.94 mm, are both acceptable for spinal indications.
However, when using the Universal AIR automatic image registration, other benefits were noted. Significant outliers in accuracy values were noticed with the manual Surface Matching registration. The analysis of these cases revealed that user errors can occur leading to wrong level registration or incorrect point acquisition. This translates into failed or inaccurate registration. The exclusion of these cases from the analysis resulted in more consistent data, but still the distribution around the mean was narrower for the automatic registration.
Overall, we found that the use of Universal AIR together with the O-arm scanner was feasible. In some cases, insufficient image quality resulted in failed registration. However, setting the scanner mode to high definition (higher radiation dose with better image quality) allowed all registration scans to be successfully used for automatic image registration. The Universal Air matrix was easy to position in the surgical field and could rest suspended on the skin on either side of the incision. Three different sizes of the Universal AIR matrix are provided by the manufacturer, for different minimally invasive and open surgical scenarios. However, care must be taken that all the reflective spheres of Universal Air and the dynamic reference frame are visible to the IR camera of the navigation system during image acquisition. We experienced that the simultaneous detection of three different reference markers (the dynamic reference frame, the reflective spheres of the Universal Air and the radiopaque markers of the Universal Air) could be challenging, especially when the IR camera is looking “through” the O-arm gantry.
Using alternative methods for patient tracking could be a possible remedy for this difficulty. Most conventional navigation systems use a dynamic reference frame attached to the spinous process of a vertebra. The dynamic reference frame is normally shaped as a “star” with reflecting spherical markers that are recognized by an infrared camera. Although very effective for patient tracking and navigation, these reference frames are often bulky and occupy space in the vicinity of the surgical field, thereby interfering with the surgical procedure. They are also easily dislodged resulting in loss of navigational accuracy. To overcome these challenges, alternative tracking solutions have been introduced. Adhesive skin markers are used by the augmented reality surgical navigation system (ARSN) commercially known as ClarifEye (Philips, Best, The Netherlands), while Spine Mask (Stryker, Kalamazoo, MI, USA) utilizes LED lights in a frame attached to the patient’s back. Several augmented reality navigation devices use manual image superimposition [
23], while sophisticated surface feature recognition methods have been proposed in experimental models [
9,
24].
Integrated systems like the Loop X (Brainlab AG) paired with Brainlab navigation technology and the O-arm (Medtronic) paired with the StealthStation navigation technology utilize the infrared camera of the image-guided navigation system to detect reflective markers that are placed on the gantry of the scanner during the scan. The navigation software transforms the coordinate system of the 3D image to the coordinates of the patient and allows for immediate image registration and navigation without additional steps. A similar automatic registration method is used in ClarifEye, a hybrid OR-based navigation system for spine surgery where a robotic C-arm harbors the navigation cameras and registration occurs simultaneously during imaging [
10].
Universal AIR instead provides a solution where non-integrated systems can be combined without losing automatic registration functionality. While Universal AIR offers the benefit of working with many intraoperative 3D imaging devices, it requires that the Universal AIR matrix and the region of interest (e.g., spine) are displayed in the same field of view (FoV) of the scanner. This can be challenging, particularly in obese patients, as the skin level with the attached matrix and the region of interest for surgery may be too distant from one another, making the use of the Universal AIR matrix registration problematic, particularly if a scanner has a limited FoV.
It has previously been shown that CBCT is as reliable as conventional CT for breach detection in spinal fixation surgeries [
25]. Although the image quality provided by the O-arm (even when the HD mode is used) is inferior to conventional high-dose CT, the combination of Universal AIR automatic registration and O-arm scans provides excellent quality for intraoperative navigation and simultaneously making additional pose corrections superfluous. Preoperative CT scans are obtained with the patient in supine position and therefore require pose correction to accurately be used in navigation, with the patient in the prone position.
During the study, five cases of miss-matched vertebrae during Surface Matching occurred. This was particularly evident in thoracic cases where the vertebra of interest in the surgical field was incorrectly matched to an adjacent vertebrae on the preoperative CT scan. This wrong level matching and possible wrong level surgery is avoided when intraoperative imaging and automatic registration is used. In several cases where the fixation extended from C2-Th2, two separate CBCTs were obtained for instrumentation of upper cervical and upper thoracis levels. In contrast to navigation based on preoperative CT, by using automatic registration on intraoperative CBCT, accurate navigation was possible despite spine movements secondary to the instrumentation at one end of the surgical field.
Strengths and limitations
This study was performed by surgeons with a long experience of different surgical navigation systems and patient registration methods. The results therefore reflect the clinical use of the technology in experienced hands but may not translate directly to inexperienced users or surgeons under training.
Navigation systems based on dynamic reference frames have been shown to lose accuracy with distance from the dynamic reference frame [
10]. Since all measurements were performed on the same vertebra where the dynamic reference frame was attached, this phenomenon was not studied in this manuscript.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.