B超图像复杂性特征分析方法诊断脂肪肝  

Fatty liver diagnosed by B-mode ultrasonography based on the complexity analysis

作　　者：王娇[1] 罗燕[2] 李德玉[1] 林江莉[1] 汪天富[1] 彭玉兰[2] 

机构地区：[1]四川大学生物医学工程系,四川成都610065 [2]四川大学华西医院超声科,四川成都610041

出　　处：《中国医学影像技术》2006年第1期135-138,共4页Chinese Journal of Medical Imaging Technology

基　　金：四川省青年科技基金资助(05ZQ026-019)。

摘　　要：目的利用复杂性分析研究脂肪肝患者B超图像纹理改变,进而诊断脂肪肝。方法通过分析正常肝脏与脂肪肝B超图像的复杂度,近似熵和近远场灰度比特征,组成特征矢量,利用反向传播人工神经网络对脂肪肝进行计算机辅助诊断。结果用80例样本建立识别模型,用50例样本进行验证,对正常肝的识别率达到100%,脂肪肝识别率达到100%。结论复杂性分析能较好地描述脂肪肝超声图像的特征,对脂肪肝的识别有着较好的性能。Objective To diagnose fatty liver by analyzing the complexity of B-mode ultrasonic images. Methods The complexity, approximate entropy and mean intensity ratio of images were studied. Feature vector of each liver image were created with the three features. Then use back-propagation artificial neural network to classify these vectors. Results The accuracy rates were 100% for normal liver and 100% for fatty liver. Conclusion Complexity analysis could indicate the texture feature of B-mode ultrasonic images of normal liver and fatty liver successfully and it could improve the diagnosis of fatty liver.

关 键 词：脂肪肝 复杂性分析 图像识别 反向传播人工神经网络 

分 类 号：R575.5[医药卫生—消化系统] R445.1[医药卫生—内科学]