研究生: |
陳昱丞 Yu-Cheng Chen |
---|---|
論文名稱: |
用於乳癌病理影像分割與診斷的深度學習模型開發 Development of Deep learning Models for Segmentation of Histopathological Images in Breast Cancer Diagnosis |
指導教授: |
王靖維
Ching-Wei Wang |
口試委員: |
陳燕麟
Yan-Lin Chen 鄭世平 Shig-Ping Cheng 武敬和 Ching-Ho Wu 許維君 Wei-Chun Hsu 王靖維 Ching-Wei Wang |
學位類別: |
碩士 Master |
系所名稱: |
應用科技學院 - 醫學工程研究所 Graduate Institute of Biomedical Engineering |
論文出版年: | 2019 |
畢業學年度: | 107 |
語文別: | 中文 |
論文頁數: | 35 |
中文關鍵詞: | 深度學習 、卷積神經網路 、醫學影像 、乳癌診斷 、全卷積式神經網路 |
外文關鍵詞: | Deep Learning, Convolutional Neural Network, Medical Imaging, Breast cancer Diagnosis, Fully Convolutional Networks |
相關次數: | 點閱:456 下載:0 |
分享至: |
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
乳癌為女性中最為常見的癌症,主要由病理學家檢查病人之病理切片作為診斷方式。然而乳癌的人工診斷與分類十分的耗時,並且需要經驗豐富的病理學家來進行診斷,因此本研究將深度學習用於乳癌病理切片分割、良惡性組織辨識。
關於以深度學習進行乳癌的分割與組織辨識,大多數現有的研究著重於良性與惡性乳腺腫瘤的二元分類。而我們則是以四種不同生物標記(E-cadherin, P63, CK14與 CK5/6)的整張切片影像為依據,將乳癌分為正常乳腺組織、上皮增生、非典型乳腺導管癌、原位導管癌與浸潤性導管癌五個階段並在病理學家的監督下標註H\&E切片影像。我們以此由四種生物標記的切片為依據所標註出之整張H\&E切片影像來訓練深度學習卷積神經網絡(全卷積神經網路),已達辨識良惡性組織並將乳癌組織分割的成果。
我們使用訓練完成的模型進行測試,最佳的模型測試的結果經過混淆矩陣量化分析後能夠在分割乳癌組織上達到準確率0.8768、精確率0.8477、召回率0.8037和綜合評價指標0.8137。在覆蓋率量化分析上,分割乳癌組織上達到覆蓋率50\%的比例為0.9248。這項研究結果顯示深度學習模型能夠達成自動化的乳癌影像分割與良惡性組織辨識,隨著更深入的研究,也可將其於臨床上使用,對於不論是學術研究或是醫療診斷都能夠有很大的貢獻。
Breast cancer, the most common cancer in women, is primarily diagnosed with inspection of histopathological slides by pathologists. Manual diagnosis and classification of breast cancer is time consuming and needs experienced pathologists,and thus this research use deep learning for segmentation of breast cancer histopathological image and recognition of benign or malignancy tissue.
About the detection and classification of breast cancer by Deep Learning. Most of the existing works focus only on binary classification of breast tumor (benign and malignant).We used whole slide images of four different biomarkers (E-cadherin, P63, CK14, and CK5/6) to categorize breast cancer stages into five different categories:normal breast tissue, Epithelial Hyperplasia, Atypical Ductal Carcinoma, Ductal Carcinoma Insitu and Invasive Ductal Carcinoma. Based on these four biomarkers we manually annotated H\&E slide under supervision of Pathologists. By using H\&E whole slide images which are annotated based on the four biomarkers, we trained the deep learning convolutional neural network (Fully Convolutional Network) to segment breast cancer tissue and recognize benign and malignancy tissue.
Our model was able to detect and classify different tissue of breast cancer with accuracy of 0.8768,precision of 0.8477,recall of 0.8037 and F-measure of 0.8137.In the evaluation of coverage ,the ratio of segmentation result which cover 50\% of the breast cancer tissue is 0.9248. These findings suggest that deep learning models can automate the process of segmentation and the recognition between benign and malignancy tissue and with further research it can be used for clinical use .
[1] Jonathan Long, Evan Shelhamer, and Trevor Darrell. Fully convolutional networks
for semantic segmentation. IEEE Conference on Computer Vision and Pattern Recognition, pages 3431–3440, 2015.
[2] B.Stewart and C.Wild. World cancer report 2014. WHO Press, World Health Organization,2014.
[3] Y.LeCun, Y.Bengio, and G.Hinton. Deep learning. Nature, 521:436–444, 2015.
[4] S.Webb. The technology feature‘deep learning for biology’. Nature, 554:555–557, 2018.
[5] G.Litjens et al. Deep learning as a tool for increased accuracy and efficiency of
histopathological diagnosis. Scientific Reports, 6(26286), 2016.
[6] L.Cooper et al. Digital pathology: Data-intensive frontier in medical imaging. Proceedings of the IEEE, 100:991–1003, 2012.
[7] ML.Saini, C.Bouzin, and K.Saini. Digital pathology: an overview. Integrative
Molecular Medicine, 4:1–2, 2017.
[8] F.Spanhol. Breast cancer hitopathological image classification using convulutional neural networks. International Joint Conference on Neural Networks (IJCNN), pages 2560–2567, 2016.
[9] Z.Han et al. Breast cancer multi-classification from histopathological images with structured deep learning model. Scientific Reports, 7(4172), 2017.
[10] B.Kieffer, M.Babaie, S.Kalra, and H.R.Tizhoosh. Convolutional neural networks for histopathology image classification: Training vs. using pre-trained networks. Seventh International Conference on Image Processing Theory, Tools and Applications
(IPTA), Montreal, pages 1–6, 2017.
[11] et al N.Coudray. Classification and mutation prediction from non-small cell lung
cancer histopathology images using deep learning. Nature Medicine, 24:1159–1567,
2018.
[12] J.Chang, J.Yu, T.Han, H.Chang, and E.Park. A method for classifying medical images using transfer learning: A pilot study on histopathology of breast cancer. IEEE, 2017-December:1–4, 2017.
[13] A.Rossetto and W.Zhou. Deep learning for categorization of lung cancer ct images. IEEE/ACM International Conference on Connected Health: Applications, Systems
and Engineering Technologies (CHASE), Philadelphia, pages 272–273, 2017.
[14] J.Kwak and S.Hewit. Nuclear architecture analysis of prostate cancer via convolutional neural networks. IEEE access, 5:18526– 18533, 2017.
[15] et al N.Balasooriya. A shophisticated convolutional neural network model for brain tumor classification. IEEE International Conference on Industrial and Information Systems (ICIIS), Peradeniya, pages 1–5, 2017.
[16] A.Godkhindi and R.Gowda. Automated detection of polyps in ct colonography
images using deep learning algorithms in colon cancer diagnosis. IEEE International
Conference on Energy, Communication, Data Analytics and Soft Computing (ICECDS), pages 1722–1728, 2017.
[17] M.Jamaluddin, M.Fauzi, and F.Abas. Tumor detection and whole slide classification of h&e lymph node images using convolutional neural network. IEEE International Conference on Signal and Image Processing Applications (ICSIPA),Kuching, pages 90–95, 2017.
[18] A.Esteva et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature, 542:115–118, 2017.
[19] A.Cruz-Roa et al. Accurate and reproducible invasive breast cancer detection in
whole-slide images: A deep learning approach for quantifying tumor extent. Scientific
Reports, 2017.
[20] Ellis IO, Schnitt SJ, and Sastre-Garau X et al. Tumours of the breast. in: Tavassoli fa, devilee p, eds.pathology and genetics of tumors of the breast and female genital organs. World Health Organization Classification of Tumours, 4:9–112, 2003.
[21] L.Khazai and M.Rosa. Use of immunohitochemical stains in epithelial lesions of the breast cancer control. IEEE Transactions on Medical Imaging, 22:220–225, 2015.
[22] et al I.Pollanen. Current analysis approaches and performance needs for whole slide image processing in breast cancer diagnostics. IEEE International Conference on
Embedded Computer Systems: Architectures, Modeling, and Simulation, pages 319–
325, 2015.
[23] D.Liu. Application of immunohistochemistry in breast pathology. Arch Pathol Lab
Med, 138:1629–1642, 2014.
[24] I.Ellis. Intraductal proliferative lesions of the breast: morphology, associated risk and molecular biology. Modern Pathology, 23:1–7, 2010.
[25] M.Peikari, M.Gangeh, J.Zubovits, G.Clarke, and A.Martel. Triaging diagnostically
relevant regions from pathology whole slides of breast cancer: A texture based approach. IEEE Transactions on Medical Imaging, 35:307–315, 2016.
[26] Vinod Nair and Geoffrey E.Hinton. Rectified linear units improve restricted boltzmann machines. ICML’10 Proceedings of the 27th International Conference on
International Conference on Machine Learning, pages 807–814, 2010.
[27] Geoffrey E. Hinton, Nitish Srivastava, Alex Krizhevsky, Ilya Sutskever, and Ruslan R. Salakhutdinov. Improving neural networks by preventing co-adaptation of
feature detectors. arXiv preprint arXiv:1207.0580, 2012.
[28] J.Yosinski, J.Clune, Y.Bengio, and H.Lipson. How transferable are features in deep neural networks? Advances in Neural Information Processing Systems, 27:3320–
3328, 2014.
[29] M.Oquab, L.Bottou, I.Laptev, and J.Sivic. Learning and transferring mid-level image representations using convolutional neural networks. IEEE Conference on Computer Vision and Pattern Recognition, pages 1717–1724, 2014.
[30] Matthew D.Zeiler and Rob Fergus. Visualizing and understanding convolutional
networks. Computer Vision-ECCV, pages 818–833, 2014.
[31] Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for
large-scale image recognition. arXiv:1409.1556, 2014.