文章摘要

文章信息/Info

Author(s):: LI Dingli; JIANG Shuangshuang; ZHOU Dan; Department of Radiology, Mianyang Central Hospital, Mianyang 621000,Sichuan,China

Keywords:: Computed tomograhp; Image features; High precision; Regional division; Dynamic adaptive; Region growing

摘要:: 传统CT图像分割在进行最为重要的边缘分割时，复杂空洞和非血管区域特征运用单一约束条件，缺少持续性操作过程，降低了分割精度。提出一种约束持续的CT图像特征的高精度区域分割方法，通过灰度腐蚀运算和滤波处理，获取灰度直方统计结果，据此确定种子点选取原则，实现对血管的初步分割，在此基础上采用更新均值和种子点持续分割作为约束条件，更新均值和种子点持续分割直至无满足灰度区间像素点，实现对CT图像特征的高精度区域分割。以肝脏CT图像为例的分割实验结果说明，所提方法可降低肝脏分割时产生的空洞和非血管信息的提取概率，相较于手工方法与传统方法，本研究所提方法的分割精度较高。

Abstract:: In traditional CT image segmentation, the most important edge segmentation, complex cavities and non-vascular region features use a single constraint condition, lack of continuous operation process, reducing the segmentation accuracy. A high-precision region segmentation method based on constrained continuous segmentation of CT image features was proposed. The gray-scale histogram was obtained by gray-scale etching operation and filtering processing. According to this method, the principle of selecting seed points was determined, and the initial segmentation of blood vessels was realized. On this basis, the updated mean and the continuous segmentation of seed points were used as constraints. Updating the mean value and the seed point segmenting continuously until the pixels did not meet the gray level interval, the high-precision region segmentation of CT image features was realized. The experimental results of liver CT image segmentation show that the proposed method can reduce the probability of extracting void and non-vascular information generated in liver segmentation. Compared with manual method and traditional method, the segmentation accuracy of the proposed method in this paper is higher.