Research On 3D Reconstruction Technology For Neurosurgical Puncture Localization

At present,neurosurgical puncture operation mainly depends on doctors’ hand feeling and experience,and the success rate of operation is limited.And due to the rapid development of computer vision technology,3D reconstruction technology has gradually become a mature and widely used technology,but this technology in clinical medicine is very limited.Therefore,this paper will study the three-dimensional reconstruction technology of binocular vision to realize the reconstruction of the three-dimensional body surface model in the local area of the patient’s head.It was applied to the puncture operation of neurosurgery,and provided the non-contact positioning and navigation for the operation with the 3D treatment model.Firstly,we need simulate the environment of the operating room,include the lighting conditions of the shadowless lamp and the sterile cloth environment with solid color as the background.By analyzing the imaging principle of camera,obtained the transformation relation of pixel coordinate system from a space point to the camera and coordinate transformation between binocular cameras.Taked the calibration checkerboard diagram to finish the calibration of camera to get internal parameters,distortion principle and the external parameter by Open CV.Secondly,corrected and epipolar matching the images.By using the open source image set,compared the advantages and disadvantages of two matching algorithms and to choose better one.To get a better disparity,need to optimize the parameters of the algorithm and add texture.Then,by using the disparity map and thr matrix Q to get point cloud.The point cloud which created by binocular vision is messing,full of chaotic and empty points and no surface information.Befor registration and reconstruction the point cloud should be filtered and cropped to remove noise point.Then,the point cloud should be calculated the normal vector,reconstruct the Poisson surface and delete the redundant surface to obtain a.stl file which can be edited by 3D modeling software.At the same time,the cropped point cloud can be registered to obtain the point cloud data with a larger scope or even the whole head.Finally,developed interactive software by Py Qt5 which contains taking picture modile,stereo camera calibrate module,disparity module and point cloud crop and registered module.It can be used in other computer without install another library by using this software.This software can help doctor get head’s point cloud easily.