摘要
临床上对早期脂肪肝的准确诊断具有较大难度,文中采用Wistar大鼠动物实验数据,研究早期脂肪肝的识别方法。通过提取肝脏超声射频信号的多个特征参数及假设检验得到的最佳特征矢量,再利用BP神经网络结合模糊函数对脂肪肝程度进行量化。结果表明,正常肝和轻度脂肪肝的识别率分别为97.50和94.29,假阴性率和假阳性率分别为2.5和5.71。
It＇ s difficult to diagnose mild fatty liver correctly in clinic,so the research of recognition of mild fatty liver by using animal experiment data of Wistar rat is proposed in this paper.Firstly several characteristic parameters are selected from ultrasound radiofrequency signal of livers,then hypothesis testing is used to obtain the best characteristic vector,finally BP neural network combined with Fuzzy function is used to quantify the degree of fatty livers.The results show that,the accuracy rates of classification is 97.50 and 94.29 for normal liver and mild fatty liver separately,and false negative rate and false positive rate is 2.5 and 5.71 separately.
出处
《实验科学与技术》
2012年第5期1-3,82,共4页
Experiment Science and Technology
基金
国家自然科学基金项目资助(30870715
30970781)
关键词
超声射频信号
脂肪肝
BP神经网络
模糊函数
ultrasonic radiofrequency signal
mild fatty liver
neural network
fuzzy function
