【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。更多还原

【Abstract】 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.更多还原