| Minimally invasive surgery can greatly reduce the suffering of patients,with the advantages of less bleeding,shorter recovery time,etc.Surgical robots are more and more widely used in modern surgery because of their precise operation and long-term surgical operation without errors.However,how to operate the surgical robot conveniently and quickly becomes a problem that plagues doctors.Some common human-computer interaction modes have some disadvantages in the operating room environment.In order to solve the above shortcomings,this paper proposes an eye gaze based automatic motion control algorithm for surgical robot.Two kinds of robotic arm movement modes based on eye gaze control were designed according to the actual needs of gynecological hysterectomy.The safety protection mechanism combined with hardware and software is designed to ensure that the robot arm will not cause secondary injury to the patient during operation.This paper adopts the end-to-end method based on convolutional neural network to predict the gaze direction.The doctor’s eye image is input into the gaze direction prediction model,which outputs the result of the gaze direction prediction for the doctor.After obtaining the gaze direction prediction information of the doctor,the gaze direction information is converted into the movement control command of the robot arm through the mapping relationship.In this paper,two moving modes of surgical robot are designed.The surgeon can use the movement of the eyeball to control the movement of the robot arm,reducing the difficulty for the medical staff to operate the robot arm.In the actual operation process,relying on one image alone for gaze direction prediction will bring relatively large interference.To solve this problem,this paper uses long-short-time memory network(LSTM)and attention mechanism(Attention)to predict eye gaze direction using multiframe images.The experimental results on the gynecological hysterectomy robot verify the feasibility of the proposed algorithm. |
PDF Full Text Request