Abstract: The auto-classification has a significant clinical value for breast cancer histopathological image. The conventional image classification method based on hand-crafted features is professional knowledge demanding and time-consuming. Besides, the conventional method can not extract discriminative features in most cases. To this end, this paper proposes an improved deep convolutional neural network model to realize automatic histopathological image classification with higher accuracy. Data augmentation and transfer learning are also adopted to avoid the overfitting by the limitation of training samples. The experimental results show that the recognition rate rises up to 91% and the proposed method has good robustness and generalization.

Key words: breast cancer histopathological image classification, deep leaning, convolutional neural network, transfer learning, data augmentation

摘要： 乳腺癌病理图像的自动分类具有重要的临床应用价值。基于人工提取特征的分类算法，存在需要专业领域知识、耗时费力、提取高质量特征困难等问题。为此，采用一种改进的深度卷积神经网络模型，实现了乳腺癌病理图像的自动分类；同时，利用数据增强和迁移学习方法，有效避免了深度学习模型受样本量限制时易出现的过拟合问题。实验结果表明，该方法的识别率可达到91%，且具有较好的鲁棒性和泛化性。

关键词: 乳腺癌病理图像分类, 深度学习, 卷积神经网络, 迁移学习, 数据增强