Robust Hand-Eye Calibration Workflow for Robotic Spine Surgery: Insights from a Cadaver Study

This paper reports the development and cadaveric validation of a Hand-Eye calibration workflow for an IR-camera-based spine surgical robot. The proposed workflow directly uses the data captured by the camera and robot encoders to perform and verify the hand-eye calibration. A self-reliant calibration integrity checking mechanism is devised to post-process the collected data for verifying and estimating the overall system error. A custom motion planner is devised to compute collision-free toolpaths for autonomously moving the robot to multiple data collection sites with minimal marker occlusions. An in-house validation was conducted to measure the accuracy of the system by subjecting it under various state-of-the-art hand-eye calibration methods. Out of 180 screw plans, Hourad's method produced the best accuracy with a positioning error of 0.97 (±0.16) mm and 0.67 (±0.09) degrees. A cadaver trial was conducted to evaluate the workflow-enabled system in a realistic spine surgical scenario. When performed intra-operatively, the proposed workflow resulted in a system error estimate of 1.79 mm. 16 screws were instrumented on both sides of the L3, L2, L1, T12, T11, T8, T7, and T6 vertebrae to assess the accuracy and performance of the system. 13 screws fell under the grade 0 breach category when it was subjected to Rampersaud classification. The results of the self-reliant calibration integrity checking mechanism were in-line with the results observed in the cadaver trial. It served to be a reliable quality control routine for verifying hand-eye calibration and estimating the system's accuracy.