论文部分内容阅读
目的:利用PCA和SVM对新疆哈萨克族食管癌X射线图像进行特征提取、特征选择及分类研究。方法:利用基于灰度共生矩阵的纹理特征和小波变换的频域特征提取法,提出将ROC曲线面积选择法和主成分分析法相结合的两步式特征选择法,利用Bayes和SVM分类器对图像进行分类以验证所提取特征的分类能力。结果:AUC>0.7的特征经主成分分析后输入到SVM分类器和Bayes分类器中得到的分类准确率和AUC值最高,分别为91%和85%、0.945和0.915。结论:SVM具有较好的分类性能,两步式特征选择法能有效地消除特征之间的共线性,极大提高了特征的分类能力,本研究有望提高新疆哈萨克族食管癌CAD系统的整体性能。
OBJECTIVE: To characterize and characterize X-ray images of Kazakh Kazakh esophageal cancer using PCA and SVM. Methods: By using the texture feature based on gray level co-occurrence matrix and the frequency-domain feature extraction method based on wavelet transform, a two-step feature selection method based on ROC curve area selection and principal component analysis was proposed. Bayes and SVM classifiers Categorize to verify the classification ability of the extracted features. Results: The features of AUC> 0.7 obtained the highest classification accuracy and AUC values of 91% and 85%, 0.945 and 0.915 respectively after being input into SVM classifier and Bayes classifier by principal component analysis. Conclusion: SVM has good classification performance. The two-step feature selection method can effectively eliminate the collinearity between features and greatly improve the classification ability of features. This study is expected to improve the overall performance of Kazakh Kazakh esophageal CAD system .