论文部分内容阅读
目的探讨数据挖掘技术在新疆肝包虫病分型中的应用。方法提取肝包虫病CT图像的灰度-梯度共生矩阵(GGCM)和灰度共生矩阵(GLCM)特征,应用主成分分析法对各纹理特征及混合特征分别进行降维,采用支持向量机(SVM)分类器、决策树C4.5分类器、Logistic回归分类器对降维后的特征进行分类,最后对各分类模型进行受试者工作特性(ROC)曲线分析及参数评估。结果 SVM分类器对不同纹理特征下3种肝脏CT图像(单囊型、多囊型肝包虫病和正常肝脏)分类效果都明显优于决策树C4.5分类器和Logistic回归分类器。综合特征分类结果要明显优于单一特征分类结果;GGCM特征对综合分类结果的分类贡献率要高于GLCM特征。结论将SVM分类器应用于新疆肝包虫病CT图像的分型中具有一定分类优势,为肝包虫病影像学诊断提供了一定的依据,也为后期新疆肝包虫病计算机辅助诊断系统的研发奠定基础。
Objective To explore the application of data mining in the classification of hepatic hydatid disease in Xinjiang. Methods Grapheme-gradient co-occurrence matrix (GGCM) and gray level co-occurrence matrix (GLCM) features of CT images of hepatic hydatid disease were extracted. Principal component analysis (PCA) was used to reduce the dimensionality of each texture feature and mixed feature separately. (SVM) classifier, C4.5 classifier and Logistic regression classifier, and classify the dimensionality reduction features. Finally, the ROC curve analysis and parameter evaluation of each classification model are carried out. Results SVM classifier had better classification accuracy than C4.5 classifier and Logistic regression classifier for three kinds of liver CT images (unicystic, polycystic liver disease and normal liver) under different texture features. The results of comprehensive feature classification are obviously better than those of single feature classification. The contribution of GGCM features to the classification results is higher than that of GLCM. Conclusion The application of SVM classifier in the classification of hepatic hydatid disease in CT image classification has a certain classification advantage, which provides a basis for the imaging diagnosis of hepatic hydatid disease, but also for the later computer-aided diagnosis system of liver hydatid disease in Xinjiang R & D laid the foundation.