基于快速最小距离树的肝脏图像非刚性配准

Nonrigid Registration Based on Minimum Spanning Tree in Liver Image

为精确诊疗肝脏疾病,针对CT肝脏图像中呈现出与相邻器官灰度值相似、边界不清晰,仅靠灰度水平法难获得理想的分割结果的问题,采用快速最小距离树度量多维特征,构造了Renyiα-entropy能量函数,成功完成了atlas和待分割图像之间的非刚性配准,实现了肝脏的自动精准分割.该方法的分割精度较传统互信息法明显提高,较传统Prim算法分割速度更快.

Segmentation of liver from CT image is difficult to obtain ideal results only by grey level method,due to the similarity of its gray scale with the adjacent organs and the unclear boundaries.To diagnose and treat liver illness accurately,Renyiα-entropy energy function was constructed,which employed minimum spanning tree to measure multidimensional features.Nonrigid registration between the atlas and unsegmented image was thus successfully completed and automatic liver segmentation was realized.The segmentation accuracy of this method is significantly improved compared with the conventional mutual information method,and the speed is faster than traditional Prim′s algorithm.