Fundus Surgical Behavior Reproduction of Robot Based on Imitation Learning

In view of the shortage of experts in fundus surgery, a method to acquire and learn the doctor's operation skills of fundus surgery is proposed. Firstly, a surgical operation acquisition platform is constructed based on the depth camera, and the surgical operation trajectory is collected. Secondly, according to the operation intention of doctors in different stages of operation, the segmentation and formulation of doctors' operation skills are realized based on hidden Markov model. Thirdly, the learning method of doctor's operation is designed based on the imitation learning algorithm, and the reproduction of doctor's operation is further realized. Finally, the relevant methods are tested and verified by using RealMan RM65-B robot. Results show that the proposed imitation learning method for fundus surgical behavior can effectively realize the representation and learning of surgical operation skills, and guide the robot to complete the relevant surgical operations.