基于SVM的新疆哈萨克族食管癌医学图像特征提取及分类研究  

Feature Extraction and Classification for Xinjiang Kazak Esophageal Cancer Based on SVM

作　　者：杨芳[1] 木拉提.哈米提 严传波[1] 阿布都艾尼.库吐鲁克 孙静[1] 姚娟[2] 

机构地区：[1]新疆医科大学医学工程技术学院,乌鲁木齐830011 [2]新疆医科大学第一附属医院影像中心,乌鲁木齐830054

出　　处：《科技通报》2016年第3期53-57,共5页Bulletin of Science and Technology

基　　金：国家自然科学基金(81460281;81160182;61201125)

摘　　要：目的:利用SVM对新疆高发病哈萨克族食管癌X线医学图像进行分类研究。方法:随机选取正常食管和缩窄型食管癌X线医学图像各120张,运用灰度直方图法和灰度共生矩阵法提取图像的特征,采用Lib-SVM工具箱,在SVM类型设置上选择C-SVC,选择4种核函数,通过调整核函数的参数与C-SVC分类器的参数进行实验。结果:利用灰度直方图法提取的特征量进行分类时,线性核函数和RBF核函数的分类准确率较高,均可达92.5%;利用灰度共生矩阵法提取的特征量进行分类时,线性核函数、RBF核函数、Sigmoid核函数的分类准确率较高,均可达87.5%;利用灰度直方图特征和灰度共生矩阵特征组成的综合特征进行分类时,多项式核函数和RBF核函数的准确率较高,均可达97.5%。结论:灰度直方图特征的分类能力优于灰度共生矩阵特征;综合特征的分类能力优于单一特征的分类能力;RBF核函数的分类性能较其他核函数突出。SVM对食管癌X线医学图像具有较高的分类识别率,为新疆高发病哈萨克族食管癌的计算机辅助诊断系统的研究奠定了基础。Objective:This paper details the feature extraction and classification for X-ray images of Xinjiang kazak esophageal cancer based on SVM. Methods:We select 120 normal esophageal X-ray image and constrictive esophageal cancer X-ray image, respectively. Gray level histogram method and gray level co-occurrence matrix method are applied to extract the image features. And the feature classification ability is evaluated by C-SVC classifier, which is included in Lib-SVM tool kit. Conducting the experiment repeatedly by adjusting the parameters of kernel function and the C-SVC classifier, respectively. Results:The classification accuracy is 92.5% for linear and RBF kernel function when using the gray level histogram feature. 87.5% for linear, RBF and Sigmoid kernel function when apply the gray level co-occurrence matrix. And 97.5 for Ploynomial kernel function when adopt the comprehensive feature. Conclusion:For the classification ability of the adopted method, gray level histogram outperforms to the gray level co-occurrence, the comprehensive feature is superior to the single feature, and RBF kernel function prevails over the other three kernel function. This work proves the classification ability of the adopted method, which lays the foundation for computer diagnosis system of kazak esophageal cancer in Xinjiang Uygur autonomous region.

关 键 词：新疆哈萨克族食管癌 灰度直方图 灰度共生矩阵 特征提取 SVM分类 

分 类 号：R735.1[医药卫生—肿瘤]