The Research On Image-guided Zygomatic Implant Surgery Robot System

Zygomatic implant technology has been regarded as an alternative treatment to massive grafting surgery in the severe atrophic maxillary.However,due to the complex structure of the zygoma anatomic and the limited surgical view,the placement of the zygomatic implant is difficult with a lot of safety concerns.Moreover,the lengths of zygomatic implants are almost 5 times longer than the dental implants,and therefore a small deviation in the entry point can cause a serious deviation at the exit point.Nowadays,the assistant method with a real-time surgical navigation has been applied to reduce the risks of zygomatic implant surgeries.However,the accuracy of the complex operation is highly dependent on the experience of the surgeon.In order to avoid disadvantages of traditional surgical navigation systems,a novel surgical robot system for the zygomatic implant placement has been designed and developed.Due to the complex structure of the zygoma anatomic,the segmentation of the important structures is of great significance to improve preoperative planning efficiency.This thesis proposed an automatic segmentation algorithm for craniofacial key anatomical structures by deep learning,and intellectualizes the preoperative planning system.(1)Firstly,a weakly-supervised segmentation approach was proposed for BG region within maxillary sinus based on multi-dilated convolution and adversarial erasing.Comparison experiments with different networks showed that the accuracy of our method was significantly improved,reducing the workload of labeling data.(2)An accurate segmentation algorithm of key anatomical structures based on fully-supervised Unet neural network is proposed,which solves the problem of high precision automatic segmentation of key anatomical structures.Through the automatic segmentation algorithm,the intellectualized upgrade of the planning software by team pre-operation has been completed,which improved the planning efficiency.In this thesis,an overall framework of the robot control system is proposed,including two parts,navigation control and real-time tracking:(1)On the one hand,a conversion algorithm of the coordinate system based on optical tracking is proposed.(2)On the other hand,this thesis proposed a real-time tracking algorithm for robots,which achieved the tracking performance about 150 ms and meets the clinical usage.Based on the developed system,system experiment has been carried out in this thesis.(1)In the segmentation of key anatomical structure,the proposed weakly supervised segmentation algorithm achieved 82.7% segmentation accuracy.The proposed fully supervised segmentation Unet model achieved 96.6% segmentation accuracy of maxillary sinus sinus cavity and 88.2% segmentation accuracy of bone graft.In order to evaluate the accuracy of the robot,phantom experiments had been carried out.The angle,entry point and exit point deviation of the robotic system were 1.37±0.52 degree,0.72±0.22 mm,and 1.66±0.57 mm,respectively.Meanwhile,a comparison between the robotic and manual operation demonstrated that the use of the surgical robot system for the zygomatic implant placement can improve the accuracy of the operation.