ABSTRACT. Purpose: The purpose of this study was to explore the potential advantages of using 3D cone beam tomosynthesis imaging to guide automated robotic pedicle screw placements. Current robotic systems are only partially automated, only providing a guide tube for manual site preparation and screw insertion. One reason robotic autonomy has been limited is because the risk introduced by system inaccuracy is high. Intraoperative 3D imaging could be used to control such risk, but current 3D systems are cumbersome, interrupt the workflow, and take many minutes to obtain a single image. 3D cone beam tomosynthesis could overcome such issues. In this study, the robotic and imaging systems were used side-by-side while placing pedicle screws in a cadaver, to explore their potential integration to provide a real-time autonomous intraoperative solution.

Methods: The study was performed in a single cadaver. The imaging system and robot were placed on opposite sides of the subject. The subject was shifted slightly toward the imaging system to capture a dual-view (RAO/LAO) tomographic image. The image was transferred to the robot’s planning workstation where a neurosurgeon used the images to plan three pedicle screw placements; one each in L2, L3, and L4. The subject was shifted laterally toward the robot where the robot was hand-guided to touch 6 fiducial markers embedded in the subject (for image/robot registration) and then enabled to automatically drill pilot holes and place the three screws. A postoperative image was captured and used to compute placement accuracy.

Results: All three screws were successfully placed by the robot with no pedicle breeches. The average insertion point error was 2.5 mm. It was possible to co-locate the robot and imaging system such that an image could be captured showing both the spine and the drill bit positioned for pilot-hole creation.

Discussion: This feasibility study demonstrates that 3D cone beam tomosynthesis images can be used to plan and guide automated robotic screw placement. The errors are comparable to commercial robotic spine surgery systems, and the dose is significantly lower. The study also demonstrates that the systems can work in close proximity to each other with minimal effort required to switch from imaging to screw placement. In some cases the systems can operate simultaneously. There could be further efficiencies derived from integrating the systems.  

ABSTRACT. Methods: A CBT system operates by acquiring fast fluoro acquisitions taken in a circular tomosynthesis geometry and reconstructing a 3D volume providing near real-time images of the surgical scene. The reconstruction is done via iterative reconstruction and makes proficient use of prior information to solve an ill posed inverse problem. Prior information can be in the form of traditional regularization, model based imaging (dynamic system and temporal evolution models), density constraint based on object recognition, and rich priors based on convolutional neural networks. The system is capable of stereo tomosynthesis. IQ and dose were previousy assessed and presented at RSNA 2018.  The purpose of this presentation is to provide the current state of the technology along with current IQ and dose numbers.  IQ is assessed via quantitative metrics measured on an accreditation phantom and qualitative assessments of cadaveric specimen imaging with and without metal implants. Scatter and absorbed dose are assessed via an ion chamber placed at operator chest height and via nanodots placed on the organs of an anthropomorphic phantom.

Results: Image quality metrics were computed, with a focus on metrics relevant to imaging for spinal fusion procedures: high-density attenuation fidelity, geometric accuracy and high-contrast resolution. Absorbed Dose is 110 uSv, Scatter Dose is 20 uRem for CBT. For the first time, CBT imaging of cadaveric thoracic and cervical specimens are being presented. These images were assessed by surgeons and show acceptable clinical usability.

Conclusions: CBT performance is approaching the current state of the art 3D imaging in terms of resolution and CNR. CBT lags in image uniformity but has clinically usable images that are obtained at a fraction of the dose, with strong potential for pediatric surgical imaging.