|
|
|
| A Deep Convolutional Networks and Combination Strategy for Automated Nuclear Atypia Grading on Breast Histopathology |
| Zhou Chao1, Xu Jun1*, Luo Bo 2 |
1(School of Information and Control, Nanjing University of Information and Technology, Nanjing 210044, China)
2(Department of Histopathology, The Central Hospital of Wuhan, Affiliated Hospital of TongJi Medical College Huazhong University of Science and Technology, Wuhan 430014, China) |
|
|
|
|
Abstract Nuclear atypia is one of important factors in Nottingham Grading System (NGS) for evaluating the aggressiveness of breast cancer. The nuclear atypia is mainly manifested in change of the nuclear shape, size, texture and uneven density. However, histologic image has complicated nature that makes the automated nuclei atypia grading a pretty difficult task. In the paper we integrated deep convolutional neural networks and combination strategy for automated nuclei atypia grading. Firstly, the histologic patches with three different resolutions were cropped into same size for training three convolutional neural networks models, respectively. During the testing, a sliding window technique was employed to choose image patches and feed to the trained DCNN. Then the majority voting was used to evaluate the grade of the image under each resolution. Finally, plurality voting was employed to evaluate the score based on three different resolutions. The proposed model got 67 points in the test set, ranking the 2nd comparing with all of current methods with good performance. Moreover, the proposed approach was computationally efficient. The average computational time on each images with the resolution of ×10, ×20, ×40 were 1.2, 5.5, and 30 seconds, respectively, indicating that the proposed approach can be applied in clinical routine procedure for automated grading of nuclei atypia on histologic images.
|
|
Received: 06 May 2016
|
|
|
|
|
|
[1] Jemal A, Siegel R, Xu Jun, et al. Cancer statistics [J]. Ca Cancer Journal for Clinicians, 2012, 62(1):10-29.
[2] Chen Wanqing, Zheng Rongshou, Baade PD, et al. Cancer statistics in China [J].Ca Cancer Journal for Clinicians, 2016, 66(2):115-132.
[3] 龚磊, 徐军, 王冠皓,等. 基于多特征描述的乳腺癌肿瘤病理自动分级[J]. 计算机应用, 2015, 35(12):3570-3575.
[4] 陈佳梅, 屈爱平, 王林伟,等. 计算机图像分析挖掘乳腺癌病理预后新指标[J]. 生物物理学报, 2014, 30(7):517-530.
[5] Robbins P, Pinder S, Klerk ND, et al. Histological grading of breast carcinomas: A study of inter observer agreement [J]. Human Pathology, 1995, 26(8):873-879.
[6] Basavanhally AN, Ganesan S, Agner S, et al. Computerized image-based detection and grading of lymphocytic infiltration in HER2+ breast cancer histopathology [J]. IEEE Transactions on Biomedical Engineering, 2010, 57(3):642-653.
[7] May M. A better lens on disease [J]. Scientific American, 2010, 302(5):74-77.
[8] Bourzac K. Software: The computer will see you now [J]. Nature, 2013, 502(7473):92-94.
[9] Roux L, Racoceanu D, Loménie N, et al. Mitosis detection in breast cancer histological images An ICPR 2012 contest [J]. Journal of Pathology Informatics, 2013, 4(1):4-8.
[10] Cruz-Roa A, Basavanhally A. Mitosis detection in breast cancer pathology images by combining handcrafted and convolutional neural network features [J]. Journal of Medical Imaging, 2014, 1(3):1-8.
[11] Veta M, Diest PJV, Willems SM, et al. Assessment of algorithms for mitosis detection in breast cancer histopathology images [J]. Medical Image Analysis, 2015, 20(1):237-248.
[12] Cire?an DC, Giusti A, Gambardella LM, et al. Mitosis detection in breast cancer histology images with deep neural networks[C]//Medical Image Computing and Computer-Assisted Intervention — MICCAI 2013. Nagoya: Springer Berlin Heidelberg, 2013:411-418.
[13] Irshad H. Automated mitosis detection in histopathology using morphological and multi-channel statistics features [J]. Journal of Pathology Informatics, 2013, 4:1-10.
[14] Khan AM, Sirinukunwattana K, Rajpoot N. A global covariance descriptor for nuclear atypia scoring in breast histopathology images [J]. Surface Science, 2015, 19(5):1637.
[15] Hinton GE, Salakhutdinov RR. Reducing the dimensionality of data with neural networks [J]. Science, 2006, 313(5786):504-507.
[16] Krizhevsky A, Sutskever I, Hinton GE. Imagenet classification with deep convolutional neural networks [J]. Advances in Neural Information Processing Systems, 2012, 25(2):2012-2022.
[17] Xu Jun, Xiang Lei, Liu Qingshan, et al. Stacked sparse autoencoder (SSAE) for nuclei detection on breast cancer histopathology images [J]. IEEE Transactions on Medical Imaging, 2015, 35(1):119-130.
[18] Xu Jun, Luo Xiaofei, Wang Guanhao, et al. A deep convolutional neural network for segmenting and classifying epithelial and stromal regions in histopathological images [J]. Neurocomputing, 2016, 191:214-223.
[19] Basavanhally A, Ganesan S, Feldman M, et al. Multi-field-of-view framework for distinguishing tumor grade in ER+ breast cancer from entire histopathology slides [J]. IEEE Transactions on Biomedical Engineering, 2013, 60(8):2089 - 2099.
[20] Koushik, Jayanth. Understanding convolutional neural networks [J].Eprint Arxiv ,25 May,2016 [Epub ahead of print].
[21] Zeiler MD, Fergus R. Visualizing and understanding convolutional networks [M]// Fleet,Computer Vision-ECCV 2014. Cham: Springer International Publishing, 2014:818-833.
[22] Alpaydin E. Combining pattern classifiers: methods and algorithms[J]. IEEE Transactions on Neural Networks, 2007, 18(3):963-964.
[23] 周志华. 机器学习[M]. 北京:清华大学出版社, 2015:171-196.
[24] Jia Yangqing, Shelhamer E, Donahue J, et al. Caffe: convolutional architecture for fast feature embedding [J]. Eprint Arxiv, 2014:675-678.
[25] Chen Lu. Automated image analysis of nuclear atypia in high‐power field histopathological image [J]. Journal of Microscopy, 2015, 258(3): 233-240. |
|
|
|
