径向基人工神经网络在宫颈细胞图像识别中的应用

Application of radial basis function artificial neural network in image diagnosis of cervical cells

目的:探讨径向基(RBF)人工神经网络在宫颈细胞图像识别中的应用。方法:提取宫颈细胞和细胞核的15个形态学特征参数及12个色度学特征参数,对700个宫颈细胞按正常、低度鳞状上皮内病变(LSIL)、高度鳞状上皮内病变(HSIL)、宫颈癌进行分类识别。利用软件STATISTICA 7.0建立网络模型并训练,用VC++.NET语言调用网络。结果:RBF网络对训练集的拟合度为97.3%,对测试集的分类准确率为95.4%。在测试集中,正常细胞的识别率为96%,LSIL细胞识别率为94%,HSIL细胞识别率为100%,癌细胞识别率为88%。RBF网络输入参数的敏感度排序与细胞病理学特征基本一致。结论:RBF人工神经网络可以很好的对宫颈细胞特别是HSIL细胞进行分类识别。

Objective： To investigate the possibility of applying artificial neural network based on radial basis function （RBF） to image recognition of cervical cells. Methods： According to 15 morphologie parameters and 12 chromatic parameters of cervical cells, 700 cervical ceils were classified as normal ceils, lowgrade squamous intrsepithelial lesion （LSIL） cells, high-grade squamous intraepithelial lesion （HSIL） ceils, and cervical cancer ceils. STATISTICA 7.0 was used to establish and train the neural network model, and VC ＋＋. NET was used to call the model. Results：The goodness of fit of the neural network model in training set was 97.3%, and the classification accuracy in testing set was 95.4%. In testing set, the recognition rate was 96% in normal ceils, 94% in LSIL cells, 100% in HSIL cells, and 88% in cervical cancer cells. The sensitivity order of input parameters in the RBF artificial neural network was approximately consistent with that of characteristics of ceil pathology. Conclusion： Cervical cancer cells, especially HSIL cells, can be well recognized by RBF artificial neural networks. RBF neural network can be widely applied in computer aided diagnosis.