论文部分内容阅读
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: To classify the X-ray medical images of Kazakh patients with high incidence of esophageal cancer in Xinjiang using SVM. Methods: 120 X-ray medical images of normal esophagus and narrowing esophageal cancer were randomly selected. The features of the images were extracted by using the gray-scale histogram method and the gray-level co-occurrence matrix method. The Lib-SVM toolbox was used to select the SVM type settings C-SVC, four kinds of kernel functions are selected, and the parameters of the kernel function and the parameters of the C-SVC classifier are adjusted. Results: The accuracy of the classification of linear kernel function and RBF kernel function was high, which both reached 92.5% when using the feature extracted by the gray histogram method. When using the gray-level co-occurrence matrix method to classify the features, The classification accuracy of linear kernel function, RBF kernel function and Sigmoid kernel function are both high, which can all reach 87.5%. When using the combination of gray histogram features and gray level co-occurrence matrix features to classify, polynomial kernel function and RBF kernel The accuracy of the function is high, up to 97.5%. Conclusion: The classification ability of gray histogram features is better than that of gray level co-occurrence matrix. The classification ability of comprehensive features is superior to the single feature classification ability. The classification performance of RBF kernel function is more prominent than other kernel functions. SVM has high classification recognition rate for esophageal X-ray medical images, which lays the foundation for the study of computer-aided diagnosis system of Kazakh esophageal cancer with high incidence in Xinjiang.