基于区域增长与统一化水平集的CT肝脏图像分割

Liver segmentation in CT images based on region-growing and unified level set method

经过混合图像预处理(图像降噪、特定比例梯度滤波、非线性灰度转换和自定义二值转换)将CT图像转化为二值图像,提供良好的种子增长环境,克服传统区域增长中增长阈值设定与种子点位置选择的困难,避免过分割.区域增长只需设置少量种子点即能大致提取完整肝脏区域.通过统一化水平集优化分割结果.该水平集由图像边缘信息与区域信息共同驱动,与单图像信息驱动的水平集相比,能适应更大的气球力与更多的迭代次数,抗边缘泄漏能力强.将该方法在SLIVER07和3Dircadb提供的共40个肝脏数据集上进行验证,结果表明:相比其他多种方法,该方法所需交互时间更少,分割准确度更高.

First,hybrid image preprocessing,consisting of anisotropic filter,scale-specific gradient magnitude filter,nonlinear grayscale conversion and customized binary conversion,was employed to transform the CT image into a binary image,providing a good seed growth condition.Thus,the difficulties of threshold setting and seed initial location setting encountered in conventional region-growing were overcome,and over-segmentation was avoided.A few seeds were required when region-growing was performed to roughly extract the whole liver region.The segmentation results were refined by the unified level set method,which was driven by both the edge information and the region information of the image.The unified level set is adapted to larger balloon force and more iterations compared with the level sets driven by single information,and has strong resistance to edge leakage.A total of 40 datasets from SLIVER07 and 3Dircadb were used for method validation.Results show that,compared with other methods,the proposed method needs less interaction time and can get higher segmentation accuracy.