融合纹理特征与形状特征的病灶CT图像识别  

Lesion CT image Classification Method of Fusion Texture Features and Shape Characteristics

作　　者：高静雅[1] 张荣国[1] 赵健 刘小君[2] GAO Jing-ya;ZHANG Rong-guo;ZHAO Jian;LIU Xiao-jun(School of Computer Science and Technology,Taiyuan University of Science and Technolog030024, China;School of Mechanical Engineering,Hefei University of Technology,Hefei 230009,China)

机构地区：[1]太原科技大学计算机科学与技术学院,太原030024 [2]合肥工业大学机械工程学院,合肥230009

出　　处：《太原科技大学学报》2018年第3期165-170,共6页Journal of Taiyuan University of Science and Technology

基　　金：国家自然基金项目(51375132);晋城市科技局项目(201501004-5);校研究生科技创新项目(20151030)

摘　　要：医学图像病变区域信息复杂,使用单一纹理特征分类效果不佳,提出了一种融合纹理与形状特征的病灶图像识别方法。首先,在常用的纹理特征基础上,融入Hough变换和不变矩两个形状特征,以考虑旋转、平移等畸变带来的图像失真影响;然后,对这些非线性分布融合信息线性化处理,通过支持向量机SVM对其进行分类,得到病灶图像与正常图像的识别结果;最后,采用某医院提供的50幅CT脑图像进行纹理特征和形状特征提取,进而进行SVM分类和识别。实验结果表明,所提出的融合两类特征信息对医学图像识别方法可行,病灶图像识别准确率有所提高。Diseased area of medical images are quite complicated,classification results do not work well when only using texture features,a new lesion images recognition method with fusing texture features and shape characteristics was brought forward. First,two shape characteristics of Hough transform and moment invariant based on common texture features were integrated wiith considering some aberration factors like rotation and translation which leads to image distortion; secondly,these nonlinear distribution that fused information were put into linearly processing,classification by support vector machine SVM,and the recognition results of lesion images and normal images were got; finally,50 CT images from hospital extract texture features and shape characteristics were used,and SVM classification experiments were carried out. The results of experiments show that proposed method integrating two kinds of characteristic information is feasible to medical images recognition,and lesion images classification accuracy has improved.

关 键 词：医学图像 纹理特性 形状特征 

分 类 号：U461[机械工程—车辆工程]