基于形态学特征提取的人工神经网络在口腔鳞状细胞癌诊断中的应用

摘要

目的 基于形态学图像处理方法,应用径向基神经网络(radial basis function,RBF)寻找一种可行、便捷的方法辅助口腔鳞状细胞癌的诊断.方法 选择口腔鳞状细胞癌和口腔非癌的组织病理切片图像进行形态学方法处理,提取表述特征的向量,作为训练集训练RBF网络;另选择67帧病理图像,包含癌和非癌的病例,作为测试集观察RBF的性能.结果 在RBF网络将测试标本分类结果的分析中可以看到不同输出值分类阈值的选择对应不同的诊断敏感度和特异度.结论 训练后的RBF虽然鉴别阳性、阴性的能力不能和金标准(即病理诊断)相比,但是通过选择不同敏感度和特异度,依然能够有效辅助病理医师,提高诊断效率,发挥机器的优势.%Objective To find a feasible and convenient way to help pathologist to analyze cell features of oral squamous cell carcinoma, an artificial neural network (radial basis function) has been applied based on the morphometric image processing. Methods Some images of histopahological sections of patients suffering from oral squamous cell carcinoma and non- carcinoma epithelium were selected to train a RBF network. The network was based on a morphometric method to extract a feature vector. Another 67 images of sections including oral squamous cell carcinoma and non- carcinoma were tested by the trained network to evaluate the performance of RBF network. Result Through the analysis of the output of trained RBF network classification, different sensitivity and specificity of diagnosis was achieved by choosing different threshold value correspongingly. Conclusion RBF network is a feasible auxiliary tool in the diagnosis of oral squamous cell carcinoma, even though it can not be a precise diagnosis standard.