当前位置: X-MOL 学术Methods › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Breast cancer histopathological image classification using a hybrid deep neural network
Methods ( IF 4.8 ) Pub Date : 2020-02-01 , DOI: 10.1016/j.ymeth.2019.06.014
Rui Yan , Fei Ren , Zihao Wang , Lihua Wang , Tong Zhang , Yudong Liu , Xiaosong Rao , Chunhou Zheng , Fa Zhang

Even with the rapid advances in medical sciences, histopathological diagnosis is still considered the gold standard in diagnosing cancer. However, the complexity of histopathological images and the dramatic increase in workload make this task time consuming, and the results may be subject to pathologist subjectivity. Therefore, the development of automatic and precise histopathological image analysis methods is essential for the field. In this paper, we propose a new hybrid convolutional and recurrent deep neural network for breast cancer histopathological image classification. Based on the richer multilevel feature representation of the histopathological image patches, our method integrates the advantages of convolutional and recurrent neural networks, and the short-term and long-term spatial correlations between patches are preserved. The experimental results show that our method outperforms the state-of-the-art method with an obtained average accuracy of 91.3% for the 4-class classification task. We also release a dataset with 3771 breast cancer histopathological images to the scientific community that is now publicly available at http://ear.ict.ac.cn/?page_id=1616. Our dataset is not only the largest publicly released dataset for breast cancer histopathological image classification, but it covers as many different subclasses spanning different age groups as possible, thus providing enough data diversity to alleviate the problem of relatively low classification accuracy of benign images.

中文翻译:

使用混合深度神经网络的乳腺癌组织病理学图像分类

即使医学科学发展迅速,组织病理学诊断仍被认为是诊断癌症的金标准。然而,组织病理学图像的复杂性和工作量的急剧增加使得这项任务非常耗时,结果可能会受到病理学家主观性的影响。因此,开发自动和精确的组织病理学图像分析方法对该领域至关重要。在本文中,我们提出了一种新的混合卷积和循环深度神经网络,用于乳腺癌组织病理学图像分类。基于组织病理学图像块更丰富的多级特征表示,我们的方法结合了卷积神经网络和递归神经网络的优点,并保留了块之间的短期和长期空间相关性。实验结果表明,我们的方法优于最先进的方法,4 类分类任务的平均准确率为 91.3%。我们还向科学界发布了一个包含 3771 张乳腺癌组织病理学图像的数据集,该数据集现已在 http://ear.ict.ac.cn/?page_id=1616 上公开可用。我们的数据集不仅是公开发布的最大的乳腺癌组织病理学图像分类数据集,而且涵盖了尽可能多的跨越不同年龄组的不同子类,从而提供了足够的数据多样性来缓解良性图像分类精度相对较低的问题。我们还向科学界发布了一个包含 3771 张乳腺癌组织病理学图像的数据集,该数据集现已在 http://ear.ict.ac.cn/?page_id=1616 上公开可用。我们的数据集不仅是公开发布的最大的乳腺癌组织病理学图像分类数据集,而且涵盖了尽可能多的跨越不同年龄组的不同子类,从而提供了足够的数据多样性来缓解良性图像分类精度相对较低的问题。我们还向科学界发布了一个包含 3771 张乳腺癌组织病理学图像的数据集,该数据集现已在 http://ear.ict.ac.cn/?page_id=1616 上公开可用。我们的数据集不仅是公开发布的最大的乳腺癌组织病理学图像分类数据集,而且涵盖了尽可能多的跨越不同年龄组的不同子类,从而提供了足够的数据多样性来缓解良性图像分类精度相对较低的问题。
更新日期:2020-02-01
down
wechat
bug