Skill-Learning-Based Trajectory Planning for Robotic Vertebral Plate Cutting: Personalization Through Surgeon Technique Integration and Neural Network Prediction

Abstract

In robotic-assisted laminectomy decompression, stable and precise vertebral plate cutting remains challenging due to manual dependency and the absence of adaptive skill-learning mechanisms. This paper presents an advanced robotic vertebral plate-cutting system that leverages patient-specific anatomical variations and replicates the surgeon’s cutting technique through a trajectory parameter prediction model. A spatial mapping relationship between artificial and patient vertebrae is first established, enabling the robot to mimic surgeon-defined trajectories with high accuracy. The robotic system’s trajectory planning begins with acquiring point cloud data of the vertebral plate, which undergoes preprocessing, Non-Uniform Rational B-Splines (NURBS) fitting, and parametric discretization. Using the processed data, a spatial mapping method translates the surgeon’s cutting path to the robotic coordinate system, with simulation validating the trajectory’s adherence to surgical requirements. To further enhance the accuracy and stability of trajectory planning, a Backpropagation(BP) neural network is implemented, providing predictive modeling for trajectory parameters. The analysis and training of the neural network confirm its effectiveness in capturing complex cutting trajectories. Finally, experimental validation, involving an artificial vertebral body model and cutting trials on patient vertebrae, demonstrates the proposed method’s capability to deliver enhanced cutting precision and stability. This skill-learning-based, personalized trajectory planning approach offers significant potential for improving the safety and quality of orthopedic robotic surgeries.