107 - Experimental protocol to detect breaches in robotic-assisted pedicle screw placement.

Introduction:
Pedicle screws are currently used to provide stable and rigid fixation of the spine. Optimal and secure placement of pedicle screws is challenging with the risk of neurological complications. Several robotic solutions are proposed in combination with navigation systems to assist the surgeon during pedicle screw insertion. None of the robotic assistance systems currently allows the detection of a breach in real-time. This study aims to establish an experimental protocol to collect different types of signals and to establish an automatic stopping algorithm before the spinal canal is compromised.

Material and

Methods:
The data collection and annotations are carried out through an experimental protocol that includes (fig1):

1. A 7-degree-of-freedom robotic arm pre-certified for medical applications.
2. A self-made velocity-controlled power drill allows measuring simultaneously the
cutting torque during drillings.
3. A threaded cutting instrument embedded with a conductivity sensor at 25Hz.
4. A transparent box with a clamping device allows rigidly fixing the vertebra while immerging the vertebral canal in a saline conductive solution.
5. A camera recording the scene from inside the vertebral canal for further annotations of the collected signals.
With this configuration, we can synchronously measure the displacement of the robot (respectively, the tooltip), the conductivity of the bone, and the cutting torque while recording a video of the instrument exiting the bone. All the recorded videos are reviewed frame by frame to determine the exact time of the violation of the vertebral canal with a precision of 0.33 milliseconds. By measuring the depth (position) of the instrument, a safety zone of 2 mm around the timestamp is chosen to process the signals. The 2 mm threshold is set to allow the degree of cortex breaching assessment according to the Gertzbein-Robbins classification (stage A or B). To evaluate the methods and materials developed as well as to create a large and open database, we collected 81 fresh lumbar pigs’ vertebrae. The applied force to drill through bone is 25 N with an angular velocity of 30 rpm.

Results:
An online processing algorithm is established on the collected data to allow an automatic stop before the breach in the medullary canal. A numerical validation is performed demonstrating that the proposed algorithm predicts the breach in 100% of the cases.
To assess experimentally the performance of the proposed breach detection algorithm, 24 vertebrae are prepared and drilled under the same conditions. Following each drilling, the prepared pedicle path is tested with a soft head probe. A scan is performed to eliminate premature stops. The robot detected cortex breaching and stopped the drilling in 100% of the cases (fig2).

Conclusion and discussion:
In this study, an original and annotated database is established to allow testing of different algorithms for detection and stopping before breaching the spinal canal. The established algorithm allows a stop in 100% of the cases both in recorded data and experimentally. The goal is to test this algorithm in real conditions to increase the safety of robotic-assisted pedicle screw placement.