Minimally Invasive Surgical Navigation System Based On Stereo Vision

Surgery navigation system is the use of CT scanning a series of tools such as the patient data and reconstruction of 3d model,using registration algorithm and scanning reconstruction surgery patients in the space model of three dimensional model matching and tracking during surgery surgical instruments,calculate the space of surgical instruments,displayed in the scan on the reconstruction of 3d model,This allows doctors to observe the current position of surgical instruments on the patient’s body in real time.Compared with traditional surgery,surgical navigation system has many advantages,such as facilitating doctors to analyze cases for preoperative diagnosis,surgical planning for patients,simulated surgery,medical teaching,etc.In addition,surgical navigation system also has the advantages of high safety,short operation time,less trauma and so on.In this paper,the tracking method of surgical instruments in the surgical navigation system was deeply studied and analyzed,the selection of positioning markers on surgical instruments and the detection algorithm were studied and analyzed,the posture analysis method of surgical instruments was studied,and the posture tracking method of surgical instruments was also deeply studied.The main research contents were as follows:(1)In view of the problem that positioning markers on surgical instruments have a great impact on the pose resolution accuracy of surgical instruments,the selection of surgical positioning markers on surgical instruments is studied,the recognition algorithm of black and white lattice positioning markers is studied,and the influencing factors on the positioning accuracy of surgical instruments are analyzed experimentatively.A symmetrically based black and white lattice detection algorithm is proposed.In the algorithm,a multi-resolution and multi-stage filtering algorithm framework is used to filter most non-black and white lattice regions in advance,which significantly reduces the computational load of the algorithm and improves the running speed of the algorithm.(2)According to the rigid body characteristics of the surgical instruments,namely,the relative position of the positioning markers on the surgical instruments is known information.For the positioning markers identified during the operation,the relative position relationship can be used to optimize and solve the posture parameters of the surgical instruments,so as to further improve the accuracy.In addition,this paper uses a method to remap the positioning marker area on the surgical instrument under the pose to the imaging plane,and takes the error between the mapping to the positioning marker area on the imaging plane and the actual detected positioning marker area as the loss function,so as to optimize the accuracy of the pose estimation.(3)According to the movement characteristics of surgical instruments,lstm recurrent neural network was used to track the posture of surgical instruments.The spatial coordinates of the positioning markers on the surgical instruments were remap back to the imaging plane to determine the area of the positioning markers on the imaging plane,reduce the recognition range of the positioning markers,and improve the running speed of the algorithm.When processing large resolution images,the improvement effect of the algorithm running time was more significant.(4)System algorithm transplantation.Considering the algorithm and hardware requirements of the surgical navigation system involved in this paper,the Jetson Nano second-generation embedded development board of nvidia was selected.The development board provided multiple camera interfaces and supported Open GL ES.This paper transferred the hardware and algorithm involved to the development board and carried out development and debugging.