基于灰度共生矩阵的新疆地方性肝包虫CT图像特征提取方法

 特征提取是图像理解与分析的关键。为提取表征新疆地方性肝包虫病的CT影像特征，提出一种基于灰度共生矩阵对肝脏和包虫病灶进行特征提取的方法。首先，对肝脏CT切片图像进行归一化，利用中值滤波和直方图均衡化对肝脏及病灶区同时进行去噪和增强，从而得到更清晰的灰度图像；然后进行灰度级压缩，利用基于灰度共生矩阵的纹理特征提取方法分别提取新疆地方性单囊型、多囊型肝包虫和正常肝脏CT图像的角二阶矩、熵、惯性矩、逆差分矩及相关性的均值和标准差作为纹理特征。统计分析发现，单囊型和多囊型肝包虫CT图像在角二阶矩、熵和逆差分矩等方面存在显著差异，具有统计学意义。最后，采用Bayes判别分类，分类正确率达到93.33%。结果表明，研究采用的纹理提取方法对描述肝包虫CT图像特征具有较理想的效果，一定程度上有助于对肝包虫CT图像进行分类和检索。

CT Image Feature Extraction Using GLCM for Xinjiang Local Liver Hydatid

The feature extraction is the key of the interpretation and analysis of an image. For extracting CT imaging features of Xinjiang local Liver hydatid, an approach is proposed, which can extract liver and hydatid lesion features at the same time, by using the gray level co-occurrence matrix. First, the liver slice CT images are normalized, while removing the noise by using the median filter and enhancing the contrast of the liver and the lesion area by using histogram equalization, to obtain a clear gray image; then, its gray-scale is reduced, gray-based Symbiosis Matrix texture feature extraction methods are used to extract texture features embodied in the mean and the standard deviation of ASM, ENT, CON, IDM and CORRLN of CT images of Xinjiang local mono-hydatid cyst and multiple daughter hydatid cyst and healthy liver. After statistical analysis, marked differences are found between mono-hydatid cyst and multiple daughter hydatid cyst CT images in ASM and ENT and IDM, as statistically significant, and finally, Bayes identification and classification are carried out, with classification accuracy rate of 93.33%. The results show the effectiveness of our method to describe liver hydatid CT images characteristics, which would help to classify and retrieve liver hydatid CT images to some extent.