Optimization of Spinal Surgery Robot Path Planning Using Deep Reinforcement Learning

In spinal surgery, the accuracy and stability of path planning are critical to the success of the operation. The traditional path planning method relies on preoperative images and doctor’s experience, which has great uncertainty and risk. To solve this problem, we propose a path planning optimization method for spinal surgery robots based on deep reinforcement learning. By combining deep learning and reinforcement learning algorithms, we built an adaptive path planning model capable of automatically generating optimal surgical paths in complex spinal anatomy. Our model has been extensively tested and validated in both simulated and real surgical Settings, and the results show that the approach outperforms traditional methods in terms of path accuracy, operative time, and safety. In addition, we discuss the adaptability and scalability of the model in different spinal surgery scenarios. Before using deep Q networks in deep reinforcement learning for optimization, the average time for spinal surgery robots to plan the optimal path in 5 surgeries was 6.39 seconds. The average time for the optimized spinal surgery robot to plan the optimal path in 5 surgeries was 2.61 seconds, and the optimized speed was significantly improved. The result can provide a new technical support and solution for spinal surgery.