卵巢癌是一種非常常見的婦科惡性腫瘤，因為在早期沒有明顯的症狀，所以大部分的患者被診斷時已為晚期。近來陸續有文獻顯示抗腫瘤新生血管的標靶藥物對某些上皮性卵巢癌(Epithelial ovarian cancer, EOC) 患者的治療有效，可延長患者壽命。然而標靶藥物費用昂貴，藥物對哪些病人適用目前臨床上並無良好之生物指標可供參考。
本研究利用經過四種不同的生物標記 (biomarker) 之免疫組織化學 (Immunohistochemistry, IHC) 染色影像，結合深度學習提出一個用於 EOC 患者在使用標靶藥物 Bevacizumab 後的結果預測系統。此研究中藉由這四種不同的 IHC 染色影像分別建立了四種深度學習 (Deep learning, DL) 模型，在量化分析結果上以 DL-AIM2 模型之性能表現最為優異。除了量化分析之外還使用了無惡化存活期 (Progression-free survival)，利用Kaplan-Meier方法以及Cox比例風險模型進行分析，結果證明了 DL-AIM2 模型能夠區分使用 Bevacizumab 治療反應良好的患者和治療無效持續復發的病患(p$=$0.004, 0.003)。證實了透過使用此模型來輔助預測患者治療效果，可得知對 Bevacizumab 有效的患者並持續地保持治療並過濾掉對 Bevacizumab 無效的患者。

Ovarian cancer is a common malignant gynecological disease.
Molecular target therapy, i.e. anti-angiogenesis with bevacizumab, was found to be effective in some patients of epithelial ovarian cancer (EOC).
Careful patient selection is critical, but there is no accessible biomarkers for routine clinical use.
We conducted a hospital-based retrospective study from March 2013 to January 2021, collecting 720 tissues cores and 12 tissue microarray (TMA) from 73 patients diagnosed at EOC and PSPC and treated with bevacizumab.
TMA with four immune-related proteins and four deep learning based systems were built.
In evaluation, one of the proposed models, namely DL-AIM2, achieves high recall 0.971, precision 0.895, and AUC 0.974. In addition, Kaplan-Meier progression-free, overall survival, and Cox proportional hazards model analysis were also used. Both analyses demonstrated the model is able to distinguish patients with good treatment responses from patients with disease progression (p<0.003) and patients who were predicted to be effective by the DL model had a lower risk of recurrence than predicted to be invalid (HR=0.18, 95% C.I= 0.06-0.55, p=0.003).
This study indicates new potential precision oncology systems to predict therapy benefits in EOC and PSPC patients.

[1] scikit­learn developers. Receiver Operating Characteristic (ROC). https: //scikit-learn.org/stable/auto_examples/model_selection/plot_ roc.html#sphx-glr-auto-examples-model-selection-plot-roc-py, 2007 ­ 2020. [Online; accessed 13­July­2021].
[2] M Steven Piver, TR Baker, M Piedmonte, and AM Sandecki. Epidemiology and etiology of ovarian cancer. In Seminars in oncology, volume 18, pages 177–185, 1991.
[3] Alexander J Cortez, Patrycja Tudrej, Katarzyna A Kujawa, and Katarzyna M Lisowska. Advances in ovarian cancer therapy. Cancer chemotherapy and pharmacology, 81(1):17–38, 2018.
[4] Usha Menon. Ovarian cancer: challenges of early detection. Nature Clinical Practice Oncology, 4(9):498–499, 2007.
[5] George Pentheroudakis and Nicholas Pavlidis. Serous papillary peritoneal carcinoma: unknown primary tumour, ovarian cancer counterpart or a distinct entity? a systematic review. Critical reviews in oncology/hematology, 75(1):27–42, 2010.
[6] Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. Deep learning. nature, 521(7553):436–444, 2015.
[7] Hayit Greenspan, Bram Van Ginneken, and Ronald M Summers. Guest editorial deep learning in medical imaging: Overview and future promise of an exciting new technique. IEEE Transactions on Medical Imaging, 35(5):1153–1159, 2016.
[8] Hui Jun Lim and William Ledger. Targeted therapy in ovarian cancer. Women＇ s Health, 12(3):363–378, 2016.
[9] Curzio Rüegg. Leukocytes, inflammation, and angiogenesis in cancer: fatal attractions. Journal of leukocyte biology, 80(4):682–684, 2006.
[10] Young­Woong Kim, Xiaoxia Z West, and Tatiana V Byzova. Inflammation and oxidative stress in angiogenesis and vascular disease. Journal of molecular medicine, 91(3):323–328, 2013.
[11] Jin­Sung Park, Il­Kug Kim, Sangyeul Han, Intae Park, Chan Kim, Jeomil Bae, Seung Ja Oh, Seungjoo Lee, Jeong Hoon Kim, Dong­Cheol Woo, et al. Normalization of tumor vessels by tie2 activation and ang2 inhibition enhances drug delivery and produces a favorable tumor microenvironment. Cancer cell, 30(6):953–967, 2016.
[12] Bradley J Monk, Heather Dalton, John H Farley, Dana M Chase, and Ivor Benjamin. Antiangiogenic agents as a maintenance strategy for advanced epithelial ovarian cancer. Critical reviews in oncology/hematology, 86(2):161–175, 2013.
[13] Junzo Hamanishi, Masaki Mandai, and Ikuo Konishi. Immune checkpoint inhibition in ovarian cancer. International immunology, 28(7):339–348, 2016.
[14] Robert A Burger, Mark F Brady, Michael A Bookman, Gini F Fleming, Bradley J Monk, Helen Huang, Robert S Mannel, Howard D Homesley, Jeffrey Fowler, Benjamin E Greer, et al. Incorporation of bevacizumab in the primary treatment of ovarian cancer. New England Journal of Medicine, 365(26):2473–2483, 2011.
[15] Amit M Oza, Adrian D Cook, Jacobus Pfisterer, Andrew Embleton, Jonathan A Ledermann, Eric Pujade­Lauraine, Gunnar Kristensen, Mark S Carey, Philip Beale, Andrés Cervantes, et al. Standard chemotherapy with or without bevacizumab for women with newly diagnosed ovarian cancer (icon7): overall survival results of a phase 3 randomised trial. The Lancet Oncology, 16(8):928–936, 2015.
[16] Diether Lambrechts, Heinz­Josef Lenz, Sanne de Haas, Peter Carmeliet, and Stefan J Scherer. Markers of response for the antiangiogenic agent bevacizumab. Journal of Clinical Oncology, 31(9):1219–1230, 2013.
[17] Alison Backen, Andrew G Renehan, Andrew R Clamp, Carlo Berzuini, Cong Zhou, Amit Oza, Selina Bannoo, Stefan J Scherer, Rosamonde E Banks, Caroline Dive, et al. The combination of circulating ang1 and tie2 levels predicts progression­free survival advantage in bevacizumab­treated patients with ovarian cancer. Clinical Cancer Research, 20(17):4549–4558, 2014.
[18] Karina Dahl Steffensen, Christine Vestergaard Madsen, Rikke Fredslund Andersen, Marianne Waldstrøm, Parvin Adimi, and Anders Jakobsen. Prognostic importance of cell­free dna in chemotherapy resistant ovarian cancer treated with bevacizumab. European journal of cancer, 50(15):2611–2618, 2014.
[19] Valentina Comunanza and Federico Bussolino. Therapy for cancer: strategy of combining anti­angiogenic and target therapies. Frontiers in cell and developmental biology, 5:101, 2017.
[20] Peter Carmeliet and Rakesh K Jain. Principles and mechanisms of vessel normalization for cancer and other angiogenic diseases. Nature reviews Drug discovery, 10(6):417–427, 2011.
[21] Jeffrey M Clarke and Herbert I Hurwitz. Understanding and targeting resistance to anti­angiogenic therapies. Journal of gastrointestinal oncology, 4(3):253–263, 2013.
[22] Francesco Crea, Lorenzo Fornaro, Guido Bocci, Lei Sun, William L Farrar, Alfredo Falcone, and Romano Danesi. Ezh2 inhibition: targeting the crossroad of tumor invasion and angiogenesis. Cancer and Metastasis Reviews, 31(3­4):753–761, 2012.
[23] Gordon C Jayson, Robert Kerbel, Lee M Ellis, and Adrian L Harris. Antiangiogenic therapy in oncology: current status and future directions. The Lancet, 388(10043): 518–529, 2016.
[24] Carlota Claussen, Achim Rody, and Lars Hanker. Treatment of recurrent epithelial ovarian cancer. Geburtshilfe und Frauenheilkunde, 80(12):1195–1204, 2020.
[25] Geoffrey E Hinton and Ruslan R Salakhutdinov. Reducing the dimensionality of data with neural networks. science, 313(5786):504–507, 2006.
[26] Kaustav Bera, Kurt A Schalper, David L Rimm, Vamsidhar Velcheti, and Anant Madabhushi. Artificial intelligence in digital pathology— new tools for diagnosis and precision oncology. Nature reviews Clinical oncology, 16(11):703–715, 2019.
[27] Jessica Vamathevan, Dominic Clark, Paul Czodrowski, Ian Dunham, Edgardo Ferran, George Lee, Bin Li, Anant Madabhushi, Parantu Shah, Michaela Spitzer, et al. Applications of machine learning in drug discovery and development. Nature Reviews Drug Discovery, 18(6):463–477, 2019.
[28] Josep Garcia, Herbert I Hurwitz, Alan B Sandler, David Miles, Robert L Coleman, Regula Deurloo, and Olivier L Chinot. Bevacizumab (avastin®) in cancer treatment: A review of 15 years of clinical experience and future outlook. Cancer Treatment Reviews, page 102017, 2020.
[29] Kate Schroder and Jurg Tschopp. The inflammasomes. cell, 140(6):821–832, 2010.
[30] Si Ming Man, Rajendra Karki, and Thirumala­Devi Kanneganti. Aim2 inflammasome in infection, cancer, and autoimmunity: role in dna sensing, inflammation, and innate immunity. European journal of immunology, 46(2):269–280, 2016.
[31] Chaelin Lee, Hien Thi Thu Do, Junhyeok Her, Yeonjae Kim, Dongkyu Seo, and Inmoo Rhee. Inflammasome as a promising therapeutic target for cancer. Life sciences, 231:116593, 2019.
[32] Larissa Ponomareva, Hongzhu Liu, Xin Duan, Eric Dickerson, Hui Shen, Ravichandran Panchanathan, and Divaker Choubey. Aim2, an ifn­inducible cytosolic dna sensor, in the development of benign prostate hyperplasia and prostate cancer. Molecular Cancer Research, 11(10):1193–1202, 2013.
[33] Chia­Ming Chang, Mong­Lien Wang, Kai­Hsi Lu, Yi­Ping Yang, Chi­Mou Juang, Peng­Hui Wang, Ren­Jun Hsu, Mu­Hsien Yu, and Cheng­Chang Chang. Integrating the dysregulated inflammasome­based molecular functionome in the malignant transformation of endometriosis­associated ovarian carcinoma. Oncotarget, 9(3): 3704–3726, 2018.
[34] Line Bjørge, Juha Hakulinen, OK Vintermyr, Hanna Jarva, Troels S Jensen, Ole Erik Iversen, and Seppo Meri. Ascitic complement system in ovarian cancer. British journal of cancer, 92(5):895–905, 2005.
[35] Kuo­Min Su, Tzu­Wei Lin, Li­Chun Liu, Yi­Pin Yang, Mong­Lien Wang, PingHsing Tsai, Peng­Hui Wang, Mu­Hsien Yu, Chia­Ming Chang, and Cheng­Chang Chang. The potential role of complement system in the progression of ovarian clear cell carcinoma inferred from the gene ontology­based immunofunctionome analysis. International journal of molecular sciences, 21(8):2824, 2020.
[36] Ruben Pio, Leticia Corrales, and John D Lambris. The role of complement in tumor growth. In Tumor microenvironment and cellular stress, pages 229–262. Springer, 2014.
[37] Jürgen Schmidhuber. Deep learning in neural networks: An overview. Neural networks, 61:85–117, 2015.
[38] Yuehao Pan, Weimin Huang, Zhiping Lin, Wanzheng Zhu, Jiayin Zhou, Jocelyn Wong, and Zhongxiang Ding. Brain tumor grading based on neural networks and convolutional neural networks. In 2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pages 699–702. IEEE, 2015.
[39] Daniel Forsberg, Erik Sjöblom, and Jeffrey L Sunshine. Detection and labeling of vertebrae in mr images using deep learning with clinical annotations as training data. Journal of digital imaging, 30(4):406–412, 2017.
[40] Mohsen Ghafoorian, Nico Karssemeijer, Tom Heskes, Inge WM van Uden, Clara I Sanchez, Geert Litjens, Frank­Erik de Leeuw, Bram van Ginneken, Elena Marchiori, and Bram Platel. Location sensitive deep convolutional neural networks for segmentation of white matter hyperintensities. Scientific Reports, 7(1):1–12, 2017.
[41] Shun Miao, Z Jane Wang, and Rui Liao. A cnn regression approach for real­time 2d/ 3d registration. IEEE transactions on medical imaging, 35(5):1352–1363, 2016.
[42] Kerstin Hammernik, Tobias Würfl, Thomas Pock, and Andreas Maier. A deep learning architecture for limited­angle computed tomography reconstruction. In Bildverarbeitung für die Medizin 2017, pages 92–97. Springer, 2017.
[43] Wentao Zhu, Chaochun Liu, Wei Fan, and Xiaohui Xie. Deeplung: Deep 3d dual path nets for automated pulmonary nodule detection and classification. In 2018 IEEE Winter Conference on Applications of Computer Vision (WACV), pages 673–681. IEEE, 2018.
[44] Rashika Mishra, Ovidiu Daescu, Patrick Leavey, Dinesh Rakheja, and Anita Sengupta. Histopathological diagnosis for viable and non­viable tumor prediction for osteosarcoma using convolutional neural network. In International Symposium on Bioinformatics Research and Applications, pages 12–23. Springer, 2017.
[45] Seung Yeon Shin, Soochahn Lee, Il Dong Yun, Sun Mi Kim, and Kyoung Mu Lee. Joint weakly and semi­supervised deep learning for localization and classification of masses in breast ultrasound images. IEEE transactions on medical imaging, 38(3): 762–774, 2018.
[46] Fuyong Xing, Yuanpu Xie, and Lin Yang. An automatic learning­based framework for robust nucleus segmentation. IEEE transactions on medical imaging, 35(2):550– 566, 2015.
[47] Olivier Simon, Rabi Yacoub, Sanjay Jain, John E Tomaszewski, and Pinaki Sarder. Multi­radial lbp features as a tool for rapid glomerular detection and assessment in whole slide histopathology images. Scientific reports, 8(1):1–11, 2018.
[48] Thomas de Bel, Meyke Hermsen, Bart Smeets, Luuk Hilbrands, Jeroen van der Laak, and Geert Litjens. Automatic segmentation of histopathological slides of renal tissue using deep learning. In Medical Imaging 2018: Digital Pathology, volume 10581, page 1058112. International Society for Optics and Photonics, 2018.
[49] Wouter Bulten, Christina A Hulsbergen­van de Kaa, Jeroen van der Laak, Geert JS Litjens, et al. Automated segmentation of epithelial tissue in prostatectomy slides using deep learning. In Medical Imaging 2018: Digital Pathology, volume 10581, page 105810S. International Society for Optics and Photonics, 2018.
[50] Korsuk Sirinukunwattana, Shan E Ahmed Raza, Yee­Wah Tsang, David RJ Snead, Ian A Cree, and Nasir M Rajpoot. Locality sensitive deep learning for detection and classification of nuclei in routine colon cancer histology images. IEEE transactions on medical imaging, 35(5):1196–1206, 2016.
[51] Angel Cruz­Roa, Hannah Gilmore, Ajay Basavanhally, Michael Feldman, Shridar Ganesan, Natalie NC Shih, John Tomaszewski, Fabio A González, and Anant Madabhushi. Accurate and reproducible invasive breast cancer detection in whole­slide images: A deep learning approach for quantifying tumor extent. Scientific reports, 7(1):1–14, 2017.
[52] Shuo Wang, Zhenyu Liu, Yu Rong, Bin Zhou, Yan Bai, Wei Wei, Meiyun Wang, Yingkun Guo, and Jie Tian. Deep learning provides a new computed tomographybased prognostic biomarker for recurrence prediction in high­grade serous ovarian cancer. Radiotherapy and Oncology, 132:171–177, 2019.
[53] Masakazu Sato, Koji Horie, Aki Hara, Yuichiro Miyamoto, Kazuko Kurihara, Kensuke Tomio, and Harushige Yokota. Application of deep learning to the classification of images from colposcopy. Oncology letters, 15(3):3518–3523, 2018.
[54] Koji Matsuo, Sanjay Purushotham, Bo Jiang, Rachel S. Mandelbaum, Tsuyoshi Takiuchi, Yan Liu, and Lynda D. Roman. Survival outcome prediction in cervical cancer: Cox models vs deep­learning model. American Journal of Obstetrics and Gynecology, 220(4):381.e1–381.e14, 2019.
[55] Jing Ke, Yiqing Shen, Yizhou Lu, Junwei Deng, Jason D Wright, Yan Zhang, Qin Huang, Dadong Wang, Naifeng Jing, Xiaoyao Liang, et al. Quantitative analysis of abnormalities in gynecologic cytopathology with deep learning. Laboratory Investigation, 101(4):513–524, 2021.
[56] Zhaowei Cai and Nuno Vasconcelos. Cascade r­cnn: high quality object detection and instance segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019.
[57] S. Ren, K. He, R. Girshick, and J. Sun. Faster r­cnn: Towards real­time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(6):1137–1149, June 2017.
[58] K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778, June 2016.
[59] Evan Shelhamer, Jonathan Long, and Trevor Darrell. Fully Convolutional Networks for Semantic Segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(4):640–651, April 2017.
[60] Z. Li, J. Zhang, T. Tan, X. Teng, X. Sun, H. Zhao, L. Liu, Y. Xiao, B. Lee, Y. Li, Q. Zhang, S. Sun, Y. Zheng, J. Yan, N. Li, Y. Hong, J. Ko, H. Jung, Y. Liu, Y. c. Chen, C. w. Wang, V. Yurovskiy, P. Maevskikh, V. Khanagha, Y. Jiang, L. Yu, Z. Liu, D. Li, P. J. Schüffler, Q. Yu, H. Chen, Y. Tang, and G. Litjens. Deep learning methods for lung cancer segmentation in whole­slide histopathology images ­ the acdc@lunghp challenge 2019. IEEE Journal of Biomedical and Health Informatics, 2020. Early Access.
[61] Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. Rethinking the inception architecture for computer vision. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2818–2826, 2016
[62] Ie­Ming Shih and Robert J Kurman. Ovarian tumorigenesis: a proposed model based on morphological and molecular genetic analysis. The American journal of pathology, 164(5):1511–1518, 2004.
[63] Andrew H Beck, Ankur R Sangoi, Samuel Leung, Robert J Marinelli, Torsten O Nielsen, Marc J Van De Vijver, Robert B West, Matt Van De Rijn, and Daphne Koller. Systematic analysis of breast cancer morphology uncovers stromal features associated with survival. Science translational medicine, 3(108):108ra113–108ra113, 2011.