【摘要】 针对乳腺癌组织病理学图像分类准确度较低的问题，结合卷积神经网络Dense Net和SENet的优点，提出SE-DenseNet网络结构用于乳腺癌组织病理学图像分类.该方法可在特征提取过程中实现对特征信息的重新标定，有效缓解梯度消失问题.结果表明，在BreaKHis数据集上的分类准确度可高达94.21%，能为临床诊断、治疗和预后提供有价值的参考.更多还原

【Abstract】 In view of the low classification accuracy of breast cancer histopathological image,combined the advantages of convolutional neural network DenseNet and SENet,this paper proposes SE-DenseNet network structure for breast cancer histopathological image classification. This method could recalibrate the feature information in the process of feature extraction and effectively alleviate the problem of gradient disappearance.The results showed that the classification accuracy on BreakHis dataset could reach 94.21%. It could provide valuable references for clinical diagnosis,treatment and prognosis.更多还原