图像融合方法在肝包虫病分型中的应用

Application of image fusion in the classification of liver hydatid disease

探讨图像融合技术在肝包虫病分型中的应用。对正常肝脏、单囊型肝包虫病、肝囊肿CT图像感兴趣区域分别使用传统的预处理和图像融合方法,对融合后的和预处理后的图像提取Tamura和灰度-梯度共生矩阵特征,通过支持向量机和BP神经网络分类模型进行分类,比较两种方法的分类准确率,并对各分类模型进行参数评估。传统预处理方法对肝囊肿CT图像Tamura和混合特征的分类效果优于图像融合方法,最佳分类准确率为98.333%;图像融合方法对单囊型肝包虫病和正常肝脏CT图像不同特征下的分类准确率均高于传统预处理方法,最佳分类准确率分别为99.167%和100%;图像融合方法不同特征不同分类器下的平均分类准确率高于传统预处理方法。将图像融合方法应用于肝包虫病CT图像的分型中具有一定的分类优势,为肝包虫病影像学诊断提供依据,也为后期研发肝包虫病计算机辅助诊断系统奠定基础。

To discuss application of image fusion technique in the classification of hepatic hydatid disease.Tamura and gray-gradient co-occurrence matrix textures were extracted from normal liver,single-cystic liver hydatid disease and hepatic cyst CT ROI images used traditional preprocessing and image fusion methods respectively.The classification accuracy of the two methods was compared by using support vector machine and BP neural network classification model,and the parameters of each classification model were evaluated.The traditional preprocessing method was superior to the image fusion method in classifying CT images of hepatic cysts with Tamura and mixed features.The best classification accuracy was 98.333%;the classification accuracy of image fusion for single-cystic liver hydatid disease and normal liver CT images was higher than that of traditional preprocessing method.The best classification accuracy was 99.167%and 100%,respectively;the average classification accuracy of image fusion method with different features and different classifiers was higher than that of traditional preprocessing methods.The application of image fusion method in the classification of hepatic hydatid CT images has certain classification advantages,which provides a basis for the imaging diagnosis of hepatic hydatid and lays a foundation for the later development of the computer-assisted diagnosis system of hepatic hydatid.