A Image Fusion Method For Surgical Navigation Based On Augmented Reality

A image fusion method for surgical navigation based on augmented reality has changedthe traditional operation pattern, that not only can relieve the pain of the patients, reducesurgical risk, but also can make preoperative planning, it has a high application value inmedical training. A image fusion based on augmented reality is adding virtual lesions to thereal operation part, the purpose of the fusion is to provide more supports for the diagnosis.This paper studies two-dimensional image fusion and display of three-dimensionalimage fusion, it can increase the amount of lesions through surface of two-dimensional imagefusion and internal physiological information of three-dimensional image fusion. In thispaper, the main research contents and the results are as follows:1. Combined with the real operation process, this paper designed augmented realitysystem, mainly including four parts: reconstruction of the virtual skull model, spacepositioning of augmented reality system, reconstruction of real operation scene, virtual-realimage fusion display. Matlab toolbox for calibration is used for binocular camera calibrationto complete the space location of the augmented reality system, that is the preparation workfor augmented reality fusion part.2. Design blue mark points that are easy identificated and extracted, use the targetwindow to locate and shear on the skull, extract blue points mark, calculate and label themark points centroid coordinates, then use image matching method based on Laplacianmatrix to registrate the left and right skull image, finally use the weighted image fusionmethod and the maximum (minimum) pixel gray level image image fusion to complete theintegration of the two-dimensional surface of skull image fusion.3. Study the theory of virtual-real registration and display method, puts forward adisplay method of image fusion, Reconstruct of the virtual skull model using3D scanningskull data, reconstruct of real scenes by using operation scenes images, finally, put the virtualskull model into the operation scenes, display on the computer screen.Do the research of the system based on the experimental platform. firstly,complete thebinocular camera calibration, through experimental analysis, the system error is within0.3pixel, reach the sub-pixel accuracy; then, use image matching method based on Laplacianmatrix to registrate the left and right skull image and fuse them; finally, put the virtual skullmodel into the real operation scene and display on the computer screen. In this paper, theresearch achievements lay the foundation for the follow-up study of dynamic visualization.