| 摘 要: 为了提高乳腺癌病理图像良恶性诊断的准确率，提出了一个基于卷积神经网络(CNN)对乳腺癌病理图像的诊断方法。利用这种方法，能够快速地对乳腺癌病理图像自动进行良恶性诊断。乳腺癌病理图像具有非常复杂的结构，利用VGG16架构的卷积神经网络对病理图像进行特征提取，利用数据增强的方法扩充数据集，使用迁移学习，将在ImageNet数据集上训练得到的权重作为该网络的初始化参数，该模型在乳腺癌数据集Breakhis上得到的准确率可以达到95%，而在经过解冻部分训练层、调整学习率等优化操作之后，分类准确率最高可以达到99%。实验结果表明，优化后的方法在乳腺癌良恶性诊断准确率方面有很大的提高。 |
| 关键词: 乳腺癌；卷积神经网络；图像分类；数据增强；迁移学习 |
|
中图分类号: TP391
文献标识码: A
|
|
| Diagnosis of Benign and Malignant Breast Cancer based on Convolutional Neural Network |
|
WANG Yang, CHEN Weiyi, MA Junshan
|
(School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China)
wyusst2021@163.com; 18018592703@163.com; 1120157572@qq.com
|
| Abstract: In order to improve the accuracy of benign and malignant diagnosis for breast cancer pathological images, this paper proposes a diagnosis method for breast cancer pathological images based on convolutional neural network (CNN). This method makes a quick and automatic benign and malignant diagnosis for breast cancer pathology images. As breast cancer pathological images have very complex structures, VGG16 (Visual Geometry Group) architecture convolutional neural network is used to extract the features of pathological images, and data enhancement is used to expand the data set. By using transfer learning, weights trained on the ImageNet data set are used as the initialized parameters of the network. The model can achieve 95% accuracy on breast cancer data set Breakhis; the classification accuracy is as high as 99% after thawing some training layers and adjusting learning rate. The experimental results show that the optimized method can greatly improve the accuracy of breast cancer benign and malignant diagnosis. |
| Keywords: breast cancer; convolutional neural network; image classification; data enhancement; transfer learning |
