一种基于CT图像的肾脏肿瘤分割算法

摘要

Because of the irregularity of shape, gray similarity and blur boundary of soft tissues in CT images, tumor segmentation is always an international difficulty. In this paper, an algorithm based on texture is proposed for kidney tumor segmentation from kidneys. The algorithm first extracts texture features of kidney and kidney tumors, and then selects effective features;finally, weighted Euclidean distance is used for segmentation. This algorithm for kidney tumor segmentation achieved good result.%由于肿瘤的形状不规则以及CT图像中软组织之间灰度相似，边界模糊，导致基于CT图像的肿瘤分割一直是一个国际难点。针对肾脏肿瘤和肾脏灰度相似导致难以分割肾脏肿瘤这一问题，提出了一种基于纹理特征的肾脏肿瘤分割算法。算法首先提取出最能体现肾脏肿瘤和肾脏差异的纹理特征，再对其进行有效性选择，最后采用加权欧氏距离分割出肾脏肿瘤，取得了良好的分割效果。