生物指標泛指患者可以被測量的診斷指標。近年來，針對帶有特定生物標記的患者，制定適合療程的『精準醫療』應運而生。精準醫療能夠讓藥物和治療更加有效並減少副作用。然而精準醫療仰賴基因定序的結果，其較高的周轉時間和測試成本阻礙了臨床的廣泛應用性。
儘管在臨床實務上並非所有的癌症患者都會做基因檢測，但一定會做病理檢查，因為病理檢查是臨床診斷癌症的標準方法。因此如能結合廣泛使用的病理切片影像以及在電腦視覺領域快速發展的深度學習方法，也許可以成為生物指標檢測的替代方案。近期的研究指出深度學習模型可以利用數位化病理切片準確預測腫瘤中的基因變異。因此，本實驗之目的在於使用數位病理切片影像和卷積神經網路預測對癌症治療有臨床意義的生物指標(包含腫瘤突變負荷量和腫瘤亞型)。
有了病理切片影像和病人的臨床檢驗結果，但是缺乏在影像級別的詳細標註，使得生物指標預測是弱監督學習問題。常見的作法是賦予同一個病人的所有影像相同的標註，我們延續這個做法，並將其視為有noisy labeling的監督學習。我們提出領域對抗師生學習網路，使用半監督學習中的師生學習網路(Teacher-Student Model)和一致性正則(consistency regularization)避免模型過擬合以及減少noisy labeling對模型訓練造成的負面影響。為解決病理切片常見的染色變化問題，我們採用機器學習領域的領域自適應方法──領域對抗神經網路(Domain Adversarial Nneural Network)，作為主幹模型。
在本篇研究中，我們採用The Cancer Genome Atlas (TCGA)資料集提供的大腸直腸癌病理切片和病人的基因檢測結果。我們成功從病理影像中預測高腫瘤突變負荷(TMB-H) ──目前美國食品藥物管理局(FDA)批准腫瘤突變負荷，可用於預測病患對於免疫治療藥物反應的指標。我們在腫瘤亞型分類的任務上和先前的研究方法做比較，實驗數據顯示我們的模型比前人的方法有10% AUROC的進步。為測試模型的泛化能力，我們在三個不同的任務上做外部資料集驗證，並對模型的半監督學習方法和領域對抗神經網路做消融實驗，實驗結果顯示領域對抗師生學習網路比基準模型擁有更好的強健性(robustness)。我們期許可以有更多的資料集做驗證，並在未來為患者提供快速有效的生物標記檢測方法。

Colorectal cancer (CRC) has been the most widespread cancer in Taiwan since 2006. How to make better diagnosis and prognosis is an important research subject. With the rise of precision medicine, doctors can provide suitable treatments for patients with specific biomarkers. However, precision treatments rely on genomic detection which is mainly acquired by genome sequencing. The excessive testing costs strictly restrict clinical usage. Motivated by the hypothesis that genetic variations of the tumor cell can result in pathological feature changes that are detectable from routine histopathology images, several prior efforts apply the deep learning method to whole-slide images to identify the genetic mutation present in the tumor.
Detecting genetic information from images is an extremely challenging weakly-supervised learning problem when only patient-level clinical outcomes are available instead of image-level annotations. Generally, we assign the same label to every patch from the same whole-slide images. However, this operation may introduce a noisy labeling problem and make the model overfit. We propose to use the teacher-student model and consistency regularization that are common in semi-supervised learning to avoid model overfitting and reduce the negative impacts of noisy labeling on model training. We further adopt a domain adversarial neural network, a representation learning method in the domain adaptation field, to address the color variability problem.
In our experiment, we use colorectal whole-slide images and patient’s clinical outcomes from the TCGA dataset, which is the largest publicly available database. We evaluate the proposed model on the classification task of tumor mutation burden and tumor subtype. We demonstrate that our proposed Domain Adversarial Teacher Student outperformed the conventional methods. We further show the generalizability of Domain Adversarial Teacher Student by training microsatellite instability, BRAF mutation, and CIMP with the TCGA dataset and validating on Harvard Tissue Microarray. We hope that our Domain Adversarial Teacher Student can be further evaluated in real-world clinical scenarios and can be used as a low-cost and reliable assay to broader patients with colorectal cancer.

Aswathy, M. A., &Jagannath, M. "Detection of breast cancer on digital histopathology images: Present status and future possibilities". Informatics in Medicine Unlocked, 8, 74–79, 2017.
Athiwaratkun, B., Finzi, M., Izmailov, P., &Wilson, A. G. "There Are Many Consistent Explanations of Unlabeled Data: Why You Should Average". ICLR, 2019.
Berthelot, D., Carlini, N., Goodfellow, I., Papernot, N., Oliver, A., &Raffel, C. "MixMatch: A Holistic Approach to Semi-Supervised Learning". ArXiv Preprint ArXiv:1905.02249, 2019.
Chang, K., Creighton, C. J., Davis, C., Donehower, L., Drummond, J., Wheeler, D., …Center, B. C. R. "The Cancer Genome Atlas Pan-Cancer analysis project". Nature Genetics, 45(10), 1113–1120, 2013.
Cheng, H.-T., Yeh, C.-F., Kuo, P.-C., Wei, A., Liu, K.-C., Ko, M.-C., …Liu, T.-L. "Selfsimilarity Student for Partial Label Histopathology Image Segmentation", 2020.
Cheng, M. L., Berger, M. F., Hyman, D. M., &Solit, D. B. "Clinical tumour sequencing for precision oncology: time for a universal strategy". Nature Reviews Cancer, 18(9), 527–528, 2018.
Coudray, N., Ocampo, P. S., Sakellaropoulos, T., Narula, N., Snuderl, M., Fenyö, D., …Tsirigos, A. "Classification and mutation prediction from non-small cell lung cancer histopathology images using deep learning.". Nature Medicine, 24(10), 1559–1567, 2018.
Ehteshami Bejnordi, B., Veta, M., vanDiest, P., vanGinneken, B., Karssemeijer, N., Litjens, G., …and the CAMELYON16 Consortium. "Diagnostic Assessment of Deep Learning Algorithms for Detection of Lymph Node Metastases in Women With Breast Cancer". JAMA, 318(22), 2199–2210, 2017.
Fu, Y., Jung, A. W., Torne, R. V., Gonzalez, S., Vöhringer, H., Shmatko, A., …Gerstung, M. "Pan-cancer computational histopathology reveals mutations, tumor composition and prognosis". Nature Cancer, 1(8), 800–810, 2020.
Ganin, Y., Ustinova, E., Ajakan, H., Germain, P., Larochelle, H., Laviolette, F., …Lempitsky, V. "Domain-Adversarial Training of Neural Networks". J. Mach. Learn. Res., 17(1), 2096–2030, 2016.
Grandvalet, Y., &Bengio, Y. "Semi-Supervised Learning by Entropy Minimization". Proceedings of the 17th International Conference on Neural Information Processing Systems, 529–536, 2004.
He, K., Fan, H., Wu, Y., Xie, S., &Girshick, R. "Momentum Contrast for Unsupervised Visual Representation Learning". ArXiv Preprint ArXiv:1911.05722, 2019.
He, K., Zhang, X., Ren, S., &Sun, J. "Deep residual learning for image recognition". Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2016-Decem, 770–778, 2016.
Hou, L., Samaras, D., Kurc, T. M., Gao, Y., Davis, J. E., &Saltz, J. H. "Patch-Based Convolutional Neural Network for Whole Slide Tissue Image Classification". Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2016-Decem, 2424–2433, 2016.
Huang, G., Liu, Z., &Weinberger, K. Q. "Densely Connected Convolutional Networks". CoRR, abs/1608.0, 2016.
Ilse, M., Tomczak, J. M., &Welling, M. "Attention-based deep multiple instance learning". 35th International Conference on Machine Learning, ICML 2018, 5(Mil), 3376–3391, 2018.
Jang, H. J., Lee, A., Kang, J., Song, I. H., &Lee, S. H. "Prediction of clinically actionable genetic alterations from colorectal cancer histopathology images using deep learning". World Journal of Gastroenterology, 26(40), 6207–6223, 2020.
Kather, J. N., Heij, L. R., Grabsch, H. I., Kooreman, L. F. S., Loeffler, C., Echle, A., …Luedde, T. "Pan-cancer image-based detection of clinically actionable genetic alterations". BioRxiv, 1–19, 2019.
Kather, J. N., Heij, L. R., Grabsch, H. I., Loeffler, C., Echle, A., Muti, H. S., …Luedde, T. "Pan-cancer image-based detection of clinically actionable genetic alterations". Nature Cancer, 1(8), 789–799, 2020.
Kather, J. N., Krisam, J., Charoentong, P., Luedde, T., Herpel, E., Weis, C. A., …Halama, N. "Predicting survival from colorectal cancer histology slides using deep learning: A retrospective multicenter study". PLoS Medicine, 16(1), 1–22, 2019.
Kather, J. N., Pearson, A. T., Halama, N., Jäger, D., Krause, J., Loosen, S. H., …Luedde, T. "Deep learning can predict microsatellite instability directly from histology in gastrointestinal cancer". Nature Medicine, 25(7), 1054–1056, 2019.
Kothari, S., Phan, J. H., &Wang, M. D. "Eliminating tissue-fold artifacts in histopathological whole-slide images for improved image-based prediction of cancer grade". Journal of Pathology Informatics, 4, 22, 2013.
Lafarge, M. W., Pluim, J. P. W., Eppenhof, K. A. J., Moeskops, P., &Veta, M. "DomainAdversarial Neural Networks to Address the Appearance Variability of Histopathology Images". In M. J.Cardoso, T.Arbel, G.Carneiro, T.Syeda-Mahmood, J. M. R. S.Tavares,
M.Moradi, …Z.Lu (Eds.), Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support (pp. 83–91), 2017.
Lee, B., &Paeng, K. "A robust and effective approach towards accurate metastasis detection and pN-stage classification in breast cancer". Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 11071 LNCS, 841–850, 2018.
Lerousseau, M., Vakalopoulou, M., Classe, M., Adam, J., Battistella, E., Carré, A., …Paragios, N. "Weakly Supervised Multiple Instance Learning Histopathological Tumor Segmentation". Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 12265 LNCS, 470–479, 2020.
Li, J., Socher, R., &Hoi, S. C. H. "DivideMix: Learning with Noisy Labels as Semisupervised Learning". International Conference on Learning Representations, 2020.
Liu, Y.-C., Ma, C.-Y., He, Z., Kuo, C.-W., Chen, K., Zhang, P., …Vajda, P. "Unbiased Teacher for Semi-Supervised Object Detection". Proceedings of the International Conference on Learning Representations (ICLR), 2021.
Marostica, E., Barber, R., Denize, T., Kohane, I. S., Signoretti, S., Golden, J. A., &Yu, K.- H. "Development of a Histopathology Informatics Pipeline for Classification and Prediction of Clinical Outcomes in Subtypes of Renal Cell Carcinoma". Clinical Cancer Research, 27(10), 2868–2878, 2021.
Nagpal, K., Foote, D., Liu, Y., Chen, P. H. C., Wulczyn, E., Tan, F., …Stumpe, M. C. "Development and validation of a deep learning algorithm for improving Gleason scoring of prostate cancer". Npj Digital Medicine, 2(1), 2019.
Nguyen, D. T., Mummadi, C. K., Ngo, T. P. N., Nguyen, T. H. P., Beggel, L., &Brox, T. "SELF: Learning to Filter Noisy Labels with Self-Ensembling". International Conference on Learning Representations, 2020.
Rusch, M., Nakitandwe, J., Shurtleff, S., Newman, S., Zhang, Z., Edmonson, M. N., …Zhang, J. "Clinical cancer genomic profiling by three-platform sequencing of whole genome, whole exome and transcriptome". Nature Communications, 9(1), 3962, 2018.
Saito, K., Watanabe, K., Ushiku, Y., &Harada, T. "Maximum Classifier Discrepancy for Unsupervised Domain Adaptation". CoRR, abs/1712.02560, 2017.
Sohn, K., Berthelot, D., Li, C.-L., Zhang, Z., Carlini, N., Cubuk, E. D., …Raffel, C. "FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence". ArXiv Preprint ArXiv:2001.07685, 2020.
Tarvainen, A., &Valpola, H. "Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results". In I.Guyon, U.VLuxburg, S.Bengio, H.Wallach, R.Fergus, S.Vishwanathan, &R.Garnett (Eds.), Advances in Neural Information Processing Systems (Vol. 30), 2017.
Tellez, D., Litjens, G., Bándi, P., Bulten, W., Bokhorst, J. M., Ciompi, F., &van derLaak, J. "Quantifying the effects of data augmentation and stain color normalization in convolutional neural networks for computational pathology". Medical Image Analysis, 58, 2019.
Yao, Y., &Dai, W. "Genomic Instability and Cancer". Journal of Carcinogenesis & Mutagenesis, 5, 1000165, 2014.
You, K., Long, M., Cao, Z., Wang, J., &Jordan, M. I. "Universal Domain Adaptation". 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2715–2724, 2019.
Zhang, C., Bengio, S., Hardt, M., Recht, B., &Vinyals, O. "Understanding deep learning requires rethinking generalization". CoRR, abs/1611.0, 2016.
Zhao, Y., Yang, F., Fang, Y., Liu, H., Zhou, N., Zhang, J., …Yao, J. "Predicting Lymph Node Metastasis Using Histopathological Images Based on Multiple Instance Learning with Deep Graph Convolution". Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 4836–4845, 2020.
Zhou, Z.-H. "A brief introduction to weakly supervised learning". National Science Review, 5(1), 44–53, 2017.