Research On Trajectory Planning Of Robot-assisted Femoral Shaft Fracture Plate Implantation Based On Deep Learning

With the advancement of our country’s industrialization process,our life has entered a new era.High-rise buildings are rising from the ground,and cars on the road are constantly flowing,the machines in the factory operate at a high speed,and the average life expectancy is greatly improved,these are all signs of social progress,but they also increase the risk of fracture to a certain extent.Femoral shaft fracture accounts for about 6%of systemic fractures,which is a common high-energy injury of lower limbs.Up to now,minimally invasive internal fixation is the preferred treatment for femoral shaft fracture.This scheme has the advantages of less trauma,fewer complications and quick recovery.However,there are some problems in the process of plate placement,such as "invisible","incorrect insertion" and"unsteady holding".Introducing robots into the operating room will solve these problems to a large extent.Therefore,the research on plate placement using robot system is carried out,in order to improve the efficiency and accuracy of plate placement.In this thesis,the locking plate is selected as the implant of bone fracture fixation,and the point cloud data of the surface of the plate is obtained with the help of a white light raster scanner,and the corresponding fixture is designed and manufactured for the plate according to the point cloud data.Using locking plate,locking plate fixture,optical three-dimensional motion capture system,femoral model and femoral model fixture,the acquisition platform of plate placement trajectory is built.Combined with the guidance of orthopedic doctors,the trajectory acquisition experiment was completed.The plate placement trajectories in various cases are obtained,for the neural network model to learn the plate placement law from it.By patching and extracting the obtained trajectory,the data set for training the neural network model is obtained.Through the analysis of the requirements of this article and the characteristics of various commonly used options of hyperparameters,some hyperparameters are determined and several levels are selected for other hyperparameters through tentative experiments.Then use orthogonal experiment and manual fine-tuning to determine all the hyperparameters of the neural network model and complete the model training.The model takes the relative position and posture before and after the plate implantation as input,and takes ten relative position and posture points representing the implantation trajectory as output,which is used to plan the required plate implantation trajectory according to the actual situation of different patients.After completing the training of the neural network model,a verification experimental platform was built to simulate the surgical environment before the plate was placed,and obtain the input of the neural network model from the experimental scene,Then the relative position and posture points on the plate placement track are obtained,which are executed by the manipulator after coordinate transformation and fitting.So as to analyze and test the output of the model and the effect of using the robotic arm to place the plate.The experimental results show:the overall change trend and position and posture fluctuation of the implantation trajectory output by the model meet the clinical needs;The model has good generalization ability and can deal with the situation that the parameters take edge values;By taking advantage of the high-precision characteristics of the robotic arm,the plate can be accurately placed into the target position and posture planned before the operation;The implantation process is safe and reliable,and there is no collision or friction between the plate and the femoral model.This thesis combines deep learning,robotics,motion capture and other related technologies to complete the planning of the plate placement trajectory,and the plate can be placed by the robotic arm according to the planned trajectory.The trajectory planning model constructed based on deep learning technology can plan the trajectory of plate placement in a targeted manner according to the actual situation of the patient,and the planned trajectory meets clinical requirements.When the robotic arm inserts the plate according to the planned trajectory,it has the characteristics of high precision,good safety,and anti fatigue.The related research in this thesis has positive significance in improving the therapeutic effect of femoral shaft fracture surgery,which lays a good foundation for the next clinical application.