Augmented navigation

Abstract

Spinal fixation procedures have the inherent risk of causing damage to vulnerable anatomical structures such as the spinal cord, nerve roots, and blood vessels. To prevent complications, several technological aids have been introduced. Surgical navigation is the most widely used, and guides the surgeon by providing the position of the surgical instruments and implants in relation to the patient anatomy based on radiographic images. Navigation can be extended by the addition of a robotic arm to replace the surgeon’s hand to increase accuracy. Another line of surgical aids is tissue sensing equipment, that recognizes different tissue types and provides a warning system built into surgical instruments. All these technologies are under continuous development and the optimal solution is yet to be found. The aim of this thesis was to study the use of Augmented Reality (AR), Virtual Reality (VR), Artificial Intelligence (AI), and tissue sensing technology in spinal navigation to improve precision and prevent surgical errors.

The aim of Paper I was to develop and validate an algorithm for automatizing the intraoperative planning of pedicle screws. An AI algorithm for automatic segmentation of the spine, and screw path suggestion was developed and evaluated. In a clinical study of advanced deformity cases, the algorithm could provide correct suggestions for 86% of all pedicles—or 95%, when cases with extremely altered anatomy were excluded. Paper II evaluated the accuracy of pedicle screw placement using a novel augmented reality surgical navigation (ARSN) system, harboring the above-developed algorithm. Twenty consecutively enrolled patients, eligible for deformity correction surgery in the thoracolumbar region, were operated on using the ARSN system. In this cohort, we found a pedicle screw placement accuracy of 94%, as measured according to the Gertzbein grading scale.

The primary goal of Paper III was to validate an extension of the ARSN system for placing pedicle screws using instrument tracking and VR. In a porcine cadaver model, it was demonstrated that VR instrument tracking could successfully be integrated with the ARSN system, resulting in pedicle devices placed within 1.7 ± 1.0 mm of the planed path. Paper IV examined the feasibility of a robot-guided system for semi-automated, minimally invasive, pedicle screw placement in a cadaveric model. Using the robotic arm, pedicle devices were placed within 0.94 ± 0.59 mm of the planned path. The use of a semi-automated surgical robot was feasible, providing a higher technical accuracy compared to non-robotic solutions. Paper V investigated the use of a tissue sensing technology, diffuse reflectance spectroscopy (DRS), for detecting the cortical bone boundary in vertebrae during pedicle screw insertions. The technology could accurately differentiate between cancellous and cortical bone and warn the surgeon before a cortical breach. Using machine learning models, the technology demonstrated a sensitivity of 98% [range: 94-100%] and a specificity of 98% [range: 91-100%].

In conclusion, several technological aids can be used to improve accuracy during spinal fixation procedures. In this thesis, the advantages of adding AR, VR, AI and tissue sensing technology to conventional navigation solutions were studied.