摘要
乳腺癌是全球范围下的女性第一大恶性肿瘤,乳腺癌的良恶性及时诊断对于癌症的早期发现及治疗具有重要的意义。传统的依靠人工诊断具有很大的主观性和局限性,并且费时费力;自动诊断工具能根据病理数据实时获取诊断结果。为提高乳腺癌病例图像自动诊断的准确率,提出了基于DenseNet网络的乳腺癌分类模型,进一步利用数据增强、迁移学习以及改进模型结构的方法,在乳腺癌病理图像数据集上建立诊断模型并进行参数优化与训练。实验结果表明,该模型在4种放大倍数下的平均识别率达到了99.20%,与以往的深度学习方法相比较,论文使用的改进DenseNet网络达到了最高的准确率。
Breast cancer is the largest malignant tumor in women worldwide.The diagnosis of benign and malignant breast cancer is great significance for the early detection and treatment of cancer.Traditional manual diagnosis has great subjectivity and limitations,and it is time-consuming and laborious.The automated diagnostic tool can obtain diagnostic results which based on pathological data in real time.In order to improve the accuracy of automatic diagnosis from the breast cancer case images,this paper proposes a classification model based on DenseNet network,further more uses the data enhancement,migration learning and improved model structure to establish a diagnostic model and optimized parameters based on the breast cancer pathological image dataset.The experimental results show that the average recognition rate of the model reached 99.2%under 4 magnifications,which achieves the highest accuracy when compared with the previous deep learning methods.
作者
刘巧利
闫航
贺鹏飞
杨信志
李彦杰
LIU Qiaoli;YAN Hang;HE Pengfei;YANG Xinzhi;LI Yanjie(College of Optoelectronic Information Science and Technology,Yantai University,Yantai 264005;Institute of Industrial Technology,Zhengzhou University,Zhengzhou 450000)
出处
《计算机与数字工程》
2019年第10期2496-2502,共7页
Computer & Digital Engineering
基金
国家自然科学基金项目(编号：61202399)
国家级大学生创新创业训练计划项目(编号：201811066004)资助
关键词
深度学习
卷积神经网络
数据增强
迁移学习
乳腺癌病理图像诊断
deep learning
convolutional neural network
data enhancement
migration learning
pathological image diagno sis of breast cancer
