鲁棒的超声图像肝硬化识别方法  

A robust method for cirrhosis recognition in ultrasound image

作　　者：雷一鸣[1] 赵希梅[1,2] 于可歆 王国栋[1] 郭卫东[4] 

机构地区：[1]青岛大学计算机科学技术学院,青岛266071 [2]山东省数字医学与计算机辅助手术重点实验室,青岛266000 [3]加州大学洛杉矶分校,洛杉矶90015 [4]青岛大学附属医院,青岛266000

出　　处：《电子测量与仪器学报》2016年第10期1460-1467,共8页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(61303079;61305045)资助项目

摘　　要：为了减少利用超声图像进行肝硬化诊断时临床医师的主观性对诊断准确性的影响,首次提出了一种利用Gabor变换和LBP特征融合的方法对肝硬化和正常肝脏进行识别。首先对原始肝脏样本分别提取其Gabor特征和LBP特征,然后将这两种特征进行融合,得到鲁棒性较强的特征,并利用C-SVM进行训练和分类。该方法克服了超声环境下肝脏图像所受的光照影响、边缘模糊,以及在尺度因素的影响下,其病变区域与正常区域的纹理用肉眼很难区分等困难,对正常肝脏和肝硬化的识别精度达到了100%,说明提出的方法可有效提高在超声环境下对肝硬化的诊断准确率,减少临床医师主观性的影响。In order to reduce the impact of subjectivity by clinician in diagnosing the liver cirrhosis using the ultrasound images, a method that utilizes the feature fusion of Gabor transform and LBP feature to recognize cirrhosis and normal liver is proposed. Firstly, Gabor feature and LBP feature of the original liver samples are extracted, then a fusion of these two features is made, and the strong robust feature is obtained. Finally, the training and classification will be accomplished by C-SVM. The proposed method has overcome some difficulties under the ultrasound circumstances such as the influences of illumination, the blurring of edges, and it is hard to distinguish the textures between the lesion region and normal region with naked eye due to the scale factor. The best recognition rate acquired by the proposed method is 100% , which indicates that the proposed method can improve the recognition rate of liver cirrhosis under the ultrasound circumstances effectively and reduce the impact of subjectivity of clinician.

关 键 词：GABOR变换 LBP特征 C—SVM 特征融合 尺度因素 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN911.73[自动化与计算机技术—计算机科学与技术]