基于改进SLNC的2D-3D医学图像配准

2D-3D Medical Image Registration based on Modified Sum of Local Normalized Correlation

介绍了一种基于改进SLNC(sum of local normalized correlation,SLNC)的2D-3D医学图像配准方法。首先对CT体积数据进行三线性插值,得到各向分辨率相同的体积数据,采用光线跟踪算法对其进行数字图像重建。针对不同位置和方向的重建图像,在灰度级压缩的基础上,用改进SLNC函数评价其与X线透视图像的相似性,利用与Brent相结合的Powell优化方法,搜索出相似性最大时的投影变换参数。将此方法用于移动数字X线投影设备——Biplanar 500采集的X线透视图像与相应CT体积数据的配准实验,得到较好的2D-3D图像配准效果。

A 2D-3D medical image registration method based on modified SLNC （sum of local normalized correlation, SLNC） function was presented. After linear interpolation, the original 3D CT data had the same resolution in all three directions, By using digitally reconstruction in different positions with the algorithm of ＂Ray Tracing＂, a series of synthetic images were achieved, which are known as digitally reconstructed radiographs （DRRs）. On the basis of gray value rescaling, the registration method depending on modified SLNC function was used as similarity measure to estimate similarity between synthetic image and radiograph. A hybrid optimization algorithm combined by Powell and Brent, was also involved to find out both position and orientation of the target. This method has already been applied to the rigid registration of fluoroscopy image gotten from mobile biplane digital fluoroscopy system Biplanar 500 and CT data, the registration result is pretty good.