文章摘要

文章信息/Info

Title:: Xinjiang Local Liver Hydatid CT Images Classification and Research based-wavelet Transform

作者:: 孔喜梅; 木拉提·哈米提 ; 严传波; 姚娟; 孙静; 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 830011

Author(s):: KONG Ximei; Murat Hamit; YAN Chuanbo; YAO Juan; SUN Jing; 1.College of Medical Engineering Technology, Xinjiang Medical University, Urumqi 830011,China; 2.Department of Radiology,First Affiliated Hospital,Xinjiang Medical University,Urumqi 830011

Keywords:: Xinjiang Local Liver Hydatid; Wavelet transformation; Region of Interest; C4.5 Decision Tree; Model evaluation

摘要:: 采用基于sym4和db4小波基两种小波变换方法，探讨对新疆地方性肝包虫CT图像的分类价值。使用sym4和db4小波两种小波基，提取感兴趣病灶区的纹理特征，并通过统计学方法筛选出特征子集，采用C4.5决策树算法构建正常肝脏和多子囊型病变肝脏CT图像的计算机分类模型，并对模型进行准确性、灵敏度和特异性的验证评估。结果显示，对正常肝脏和多子囊型肝包虫进行分类，sym4小波的识别正确率为92.5%，db4小波的识别正确率为97.5%。实验结果表明,小波变换法所提取的纹理特征对识别正常肝脏和多子囊型肝包虫CT影像有较好的意义，也为后续的基于内容的新疆地方性肝包虫病的诊断系统奠定了基础。

Abstract:: To explore classification value for Xinjiang local liver hydatid using two kinds of wavelet transform methods.This two methods consists of sym4 and db4 wavelet. Two wavelet methods were used to extract texture feature of region of interest focal zone,statistical method was used to select the optimal texture feature from the set of extracted features.The C4.5 Decision Tree was employed as a classifier.The results of C4.5 Decision Tree for sym4 and db4 wavelet analysis methods were evaluated using accuracy,sensitivity and specificity and the area under the ROC curve(AUC).The results showed that the accuracy rate of sym4 wavelet classifing normal liver and poly-cystic liver hydatid reached to 92.5%;the accuracy rate of db4 wavelet classification reached to 97.5%.The experimental results show that db4 wavelet methods is able to achieve higher classification accuracy effectiveness,it can lay a foundation which is subsequent based-content diagnostic system of Xinjiang local liver hydatid.