用于脂肪肝量化分级的B超图像特征提取  

B-Scan Ultrasound Image Feature Extraction for Quantitative Grading of Fatty Liver Severity

作　　者：闵志方[1] 宋恩民[1,2] 金人超[1] 李国宽[1] 

机构地区：[1]华中科技大学计算机科学与技术学院,武汉430074 [2]江西中医学院计算机学院,南昌330006

出　　处：《计算机辅助设计与图形学学报》2009年第6期752-757,共6页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家“八六三”高技术研究发展计划(2006AA02Z347);广东省教育部产学研结(2006D90304024)

摘　　要：为解决B超检查中医生对脂肪肝严重程度进行分级时存在的主观性强、偏差大的问题,通过提取4种用于肝脏B超图像量化分级的特征及其提取算法,为不同严重程度脂肪肝的临床诊断提供一个重要的客观参考依据.在去除图像灰度和对比度的影响后,通过高斯差分算法得到反映近场回声特性的光斑个数和大小;然后通过统计近远场图像灰度信息得到反映远场能量衰减的近远场灰度比及标准差比;最后结合已有的特征集进行最优特征向量选取.实验结果表明,文中4个特征构成的特征向量能最有效地区分正常肝和轻度、中度、重度脂肪肝.To reduce the strong subjectivity and large deviations of common existence in the clinical B-scan ultrasonic grading of fatty liver severity, 4 kinds of features and the corresponding automatic extraction methods are proposed for quantitatively grading fatty liver severities on B-scan ultrasound images. First, after eliminating the influences of gray levels and contrasts of the ultrasound images, the number and size of speckles, which represent the near-field echo characteristics, are calculated using the difference of Gaussian operator. Then, the near-field vs. far-field mean intensity ratio and the near-field vs. far-field standard deviation ratio, which represent the far-field energy attenuation, are calculated. Finally, a feature vector selection step is carried out based on the feature set. The experimental results show that the feature vector composed of the above four kinds of features can effectively discriminate among normal liver, mild, moderate and severe fatty liver.

关 键 词：超声图像 光斑特征 能量衰减特征 人工神经网络 

分 类 号：TP391.7[自动化与计算机技术—计算机应用技术]