新疆高发病肝包虫病CT图像的特征提取与分析

发布时间：2018-02-13 19:26

本文关键词： 新疆高发病肝棘球蚴病 CT图像特征提取疾病分类　出处：《新疆医科大学》2013年硕士论文　论文类型：学位论文

【摘要】：目的：对新疆高发病肝包虫病CT图像进行特征提取与特征分析，选择具有较强分类能力的特征，进一步探讨该特征在肝包虫病图像分类中的应用，为基于内容的新疆高发病肝包虫病医学图像的检索系统奠定基础。方法：使用Matlab图像处理软件，对CT图像进行预处理，改善图像的质量，，保存有效信息，删除无用信息；进而对处理后图像提取基于灰度直方图、灰度共生矩阵和柯尔莫戈洛夫复杂性的特征。使用SPSS统计分析软件，对图像特征进行最大类间距法分析和显著性分析，并且根据分析结果组成图像的综合特征；进一步使用分析所得特征对新疆高发病肝包虫病CT图像分类。结果：对新疆高发病肝包虫病CT图像灰度直方图、灰度共生矩阵和柯尔莫戈洛夫复杂性特征，使用最大类间距方法分析，结果显示，用灰度直方图特征、灰度共生矩阵特征和综合特征分类中，分类正常肝脏图像和单囊型肝包虫病图像时，分类准确率分别是81%和71%，85%和66%，91%和87%；分类正常肝脏图像和多囊型肝包虫病图像时，分类准确率分别为89%和82%，81%和72%，90%和93%；分类单囊型肝包虫病图像和多囊型肝包虫病图像时，分类准确率分别是75%和74%，75%和76%，85%和80%。对图像灰度直方图、灰度共生矩阵和柯尔莫戈洛夫复杂性特征，使用显著性方法分析，结果显示，用灰度直方图特征、灰度共生矩阵特征和综合特征分类正常肝脏图像、单囊型肝包虫病图像和多囊型肝包虫病图像时，分类准确率分别为84%、58%和77%，82%、77%和87%，96%、86%和86%。结论：将图像特征提取方法成功引入新疆高发病肝包虫病CT图像的分析中，对肝包虫病CT图像进行特征提取和特征分析并生成图像的综合特征，该特征对新疆高发病肝包虫病CT图像的分类准确率相对单一特征较高，在一定程度上满足分类需求，且特征分析结果可以进一步应用到基于内容的新疆高发病肝包虫病医学图像检索系统中，具有一定的应用价值。
[Abstract]:Objective: to extract and analyze the CT features of high incidence liver hydatid disease in Xinjiang, select the feature with strong classification ability, and discuss the application of this feature in the classification of liver hydatid disease image. Methods: Matlab image processing software was used to preprocess CT images, improve the quality of images, save effective information and delete useless information. Then, the features based on gray histogram, gray level co-occurrence matrix and Colmogorov complexity are extracted from the processed image. Using SPSS statistical analysis software, the maximum class spacing method and significance analysis are used to analyze the image features. According to the analysis results, the comprehensive features of the images are formed, and the CT image classification of high incidence liver hydatid disease in Xinjiang is further used. Results: the gray histogram of CT image of high incidence liver hydatid disease in Xinjiang is analyzed. Grey level co-occurrence matrix and Colmogorov complexity feature are analyzed by the method of maximum class spacing. The results show that, in the classification of gray histogram feature, gray level co-occurrence matrix feature and synthesis feature, When classifying normal liver images and single-cystic liver hydatidosis images, the classification accuracy was 81% and 71%, respectively, and 66% and 91% and 87%, respectively, while classifying normal liver images and polycystic hepatic hydatidosis images, The classification accuracy rates were 89% and 821% and 722% and 93%, respectively. The classification accuracy was 75% and 7475% for monocystic liver hydatidosis images and 76,85% and 80% for polycystic liver hydatidosis images, respectively. Gray level co-occurrence matrix and Colmogorov complex feature were analyzed by using significant method. The results showed that normal liver images were classified by gray histogram feature, gray level co-occurrence matrix feature and comprehensive feature. The classification accuracy of monocystic hepatic hydatidosis and polycystic hepatic hydatidosis was 84% and 77 82%, respectively. Conclusion: the method of image feature extraction was successfully introduced to the analysis of CT images of high incidence liver hydatid disease in Xinjiang. The CT image of liver hydatid disease was extracted and analyzed and the comprehensive features of the image were generated. The classification accuracy of the CT image of liver hydatid disease in Xinjiang was relatively higher than that of the single feature, which met the classification needs to some extent. The results of feature analysis can be further applied to the medical image retrieval system of high incidence liver hydatid disease in Xinjiang based on content, which has certain application value.
【学位授予单位】：新疆医科大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.41;R532.32

【参考文献】