摘要
目的:探讨人工神经网络在诊断卵泡形态质量上的价值。方法:使用MATALB软件中的神经网络工具箱(NeuralNetwork Toolbox),设计一个自组织竞争型神经网络并训练网络。网络以五个形态参数(卵泡面积、周长、最大直径、等效圆直径、似圆度)为输入,输出为健康、一般和差三种形态类型。结果:训练好的神经网络可以准确的对卵泡形态分类。结论:人工神经网络在诊断卵泡形态质量上有很好的应用前景。
Objective：To investigate the effect of artificial neural networks on measuring the parameters of morphology of pathology.Methods：A self-organizing competitive neural network is designed and trained using Neural Network Toolbox in MATALB.The input to this network are five morphological parameters（follicle area,perimeter,major diameter,equivalent circle diameter and similar circle degree）,and the output is good,general or bad.Results：Trained neural networks can form an accurate classification of the follicles.Conclusions：Diagnosis the follicular morphology using artificial neural networks shows a potential prospect of application.
出处
《中国医学物理学杂志》
CSCD
2010年第5期2166-2168,共3页
Chinese Journal of Medical Physics
关键词
卵泡形态质量
图像处理工具箱
自组织竞争网络
quality of follicle morphology
image processing toolbox
self organizing competitive network
