Elsevier

Journal of Clinical Neuroscience

Volume 72, February 2020, Pages 350-356
Journal of Clinical Neuroscience

Clinical study
Machine vision augmented reality for pedicle screw insertion during spine surgery

https://doi.org/10.1016/j.jocn.2019.12.067Get rights and content

Abstract

Implementing pedicle safe zones with augmented reality has the potential to improve operating room workflow during pedicle screw insertion. These safe zones will allow for image guidance when tracked instruments are unavailable. Using the correct screw trajectory as a reference angle for a successful screw insertion, we will determine the angles which lead to medial, lateral, superior and inferior breaches. These breaches serve as the boundaries of the safe zones. Measuring safe zones from the view of the surgical site and comparing to the radiological view will further understand the visual relationship between the radiological scans and the surgical site. Safe zones were measured on a spine phantom and were then replicated on patients. It was found that the largest causes for variance was between each of the camera views and the radiological views. The differences between the left and right cameras were insignificant. Overall, the camera angles appeared to be larger than the radiological angles. The magnification effect found in the surgical site result in an increased level of angle sensitivity for pedicle screw insertion techniques. By designing a virtual road map on top of the surgical site directly using tracked tools, the magnification effect is already taken into consideration during surgery. Future initiatives include the use of an augmented reality headset.

Introduction

There are several spine related pathologies which require neurosurgical intervention. These include disorders which are non-traumatic and degenerative as well as traumatic injuries. Treatment of the spine frequently requires the use of implants such as pedicle screws or rods to be inserted into the anatomy. During procedures such as spinal decompression and fusion, implants are required to correct for deformities and stabilize the spine. An improper screw insertion can create a breach within the spinal column or the surrounding region of the vertebrae. When this procedure is done free-hand the incident rate of breaching the spinal column ranges from 3 to 55% for the thoracic spine, and 5–41% for lumbar spine using standard acceptable free-hand technique [1], [2], [3], [4]. Resulting complications include acute neural and vascular injury, longer term mechanical failure, life-or-limb complications or a required additional surgery for revision.

Alternative to freehand surgery involves the use of computer assisted navigation (CAN). CAN is a tool currently used during spine or cranial cases in the operating room (OR). It is most often used in cranial procedures, but can be used for multiple spine procedures, including pedicle screw insertion [5]. Before navigation can be used, medical scans are required from either a CT or MRI. This information can be presented in the navigation system in the form of 2D or 3D guidance [5]. The system also requires a method of registration between the surface of the surgical site and the medical scans. This technology assists surgeons through tracking the positions of tools using infrared (IR) tracking. It then displays views of the structures lying underneath the tool tip to further inform surgeons of the patient anatomy currently being treated.

By using surgical navigation, surgeons can see key structures are underneath the surface and take preventative measures to ensure the best patient outcomes. The accuracy of the position of the subsurface structures is correlated to the accuracy of the IR tracking device and the registration accuracy between the patient’s medical scans and the surface of the surgical site. Several studies have been published showing that pedicle screw placement accuracy using 2D navigation is higher than when a free hand technique or preoperative CT scan navigation are used [6], [7], [8], [9]. Surgical navigation has been shown to reduce breach rates to less than 10% [10].

A key technology in CAN is the embedded (IR) tracking technology. Each tracked tool has a unique orientation of IR reflective markers to distinguish themselves from other tracked tools. The system is also aware of the location of the tool tip with respect to the markers [11]. Since the exact orientation of these markers is known, the system can determine the position and rotation in 3D of the tool to sub millimeter accuracy. If the IR markers attached to the tools are not in the view of the IR tracking device, the system is unable to localize the position and orientation of the tools. At times during the operation, surgeons may require tools which do not have IR markers, making them untraceable to the IR tracking system.

While there is an entire suite of tracked tools available in many operating rooms in developed countries, there are still multiple issues which must be addressed. One persistent issue in these devices is the lack of cross compatibility between systems [12]. Although an entire system may have many tools to track, if the surgeon decides to use a tracked tool belonging to another system (be it another screwdriver, screw, or pedicle finder), they will not be able to provide the same tracking capabilities if any. Some navigation systems such as the Medtronic StealthStation have IR marker arrays in their suite that are transferrable between tools. If an IR marker array alone is transferred from one tool to another tool which does not correspond to the same associated tool tip, the tool trajectory shown on the screen could be incorrect. This practice is inherently dangerous and does not necessarily maintain the same level of accuracy and safety [12]. Another issue involves using tools which have no markers at all – such as, for example, disposable biopsy needles. During cases, navigation of these disposables which require accurate insertion can be quite beneficial. However, disposables are rarely tracked because of the costs associated with the designs of these tools – especially considering they are single use items. Because of the need for recalling a trajectory to the sub-millimeter level, it is likely that the placement of untracked pedicle screws or biopsy needles will not be accurate to the same degree. Since tools are not cross compatible with other systems, it is also possible that surgeons must use tools which cannot be tracked after pedicle cannulation. The issue of maintaining insertion accuracy with untracked tools is an issue which must be addressed.

Despite the accuracy of CAN, only 11% of surgeons use it routinely for spine surgery [13]. Spine surgeons acknowledge that the technology is useful, but it is not considered easy to use and poorly integration into the surgical workflow [13]. The workflow becomes disjunct where the information which is presented in the surgical site is not easily interpreted in relation to the radiological view of the trajectory. Fig. 1 which demonstrates the visual disconnect. Surgeons must spatially connect the information presented on the computer screen to the real surgical site itself. Drawing spatial parallels in trajectories and then repeating them from memory on patient anatomy can cause error in recreating the screw trajectory.

To our knowledge, this is the first study which aims to maintain the precision and accuracy of surgical navigation when tracking is unavailable tools regardless of manufacturer or IR marker configuration. Safe zones are superimposed onto the surgical site using video see-through to assist in visualization of target trajectories with respect to patient anatomy. Moreover, the method is not subject to the pitfalls of using non-validated array attachments seen in some systems, potentially allowing for improved implant insertion accuracy.

Section snippets

Methods

In this paper, we propose a solution using augmented reality (AR) to display the correct trajectories made from tracked tools so that recreating them from memory is removed from the OR workflow. Implementing an AR roadmap removes the need for recalling the trajectory while using untracked tools. Using a novel machine vision image guided system, an overhead view from the two embedded cameras provide a method of displaying the surface of the surgical site in real time. Fig. 2 below displays where

Statistical analysis

When viewing Fig. 5, Fig. 6, Fig. 7, it is apparent that the angles viewed by the overhead cameras are relatively large in comparison to the radiological angles. This observation is further confirmed by the ANOVA and Tukey tests. The mean angles were as follows: left camera angles were 45° ± 48°, right camera angles were 48° ± 52°, and the radiological angles were 14° ± 8°, with statistically significant differences in ANOVA testing (p = 0.05). Following ANOVA, Tukey testing shows that the

Limitations of the study

In the case of the 7D Surgical MvIGS system, the placement of the overhead IR tracking system also governs the placement and orientation of the visual cameras. Since the orientation varies slightly between cases, some angles can appear to be more exaggerated than others despite being on the same spine level. This is further confirmed due to the compensation between the cameras. To limit the rotational error of the surgical navigation head, surgeons align one of the cameras to be more caudal to

Conclusion

An angle measurement system was used to characterize medial, lateral, superior and inferior breaches. The medial and lateral breach trajectories were used to physically display the start and end of the safe zone in the axial views. The superior and inferior breach trajectories were used analogously in the sagittal views. Throughout all levels of the spine, there appears to be a magnification effect between the safe zones in the radiological views and the angles observed in the left and right

References (14)

There are more references available in the full text version of this article.

Cited by (18)

  • Comparison of novel machine vision spinal image guidance system with existing 3D fluoroscopy-based navigation system: a randomized prospective study

    2022, Spine Journal
    Citation Excerpt :

    Individual vertebras can be registered to account for relative movement of spinal levels between surgery and the CT, but exposed bone is required for registration, limiting utility in minimally invasive procedures. Initial non-randomized reports using FLASH have demonstrated pedicle screw accuracy rates equivalent to contemporary 3D fluoroscopic IGS, whilst providing a reduction in time and radiation by eliminating the need for intraoperative 3D fluoroscopy [4,9,11–14]. A prospective clinical study of 171 craniospinal surgical procedures compared FLASH to 2 different 3D fluoroscopy-based navigation systems showed no significant differences between breach rates with decreased registration and setup time (41s vs. 258s and 794s) [11].

View all citing articles on Scopus
View full text