臺灣博碩士論文加值系統

醫學影像分割是醫學影像分析中基本的挑戰。在醫療影像於深度學習的應用中，最主要關注的問題就是缺少足夠的已標註影像。由於其標註過程需要專業醫師的知識與經驗，並且醫療影像中經常存在太多物件(例如:細胞)，這些皆導致大量的標註工作與成本。另一個關注的問題則是資料不平衡，這是醫學影像中經常會碰到的障礙。在淋巴球偵測的應用中，存在一重要的淋巴球亞群(例如:漿細胞)相較於其他細胞的數量少了非常多，這使的訓練會偏向占多數樣本族群的那一類。然而，傳統的標註策略，例如:主動學習，對於找到足夠的少數樣本族群是沒效率的。因此，本研究採用一種低成本的人工註釋方法，且可以在極端不平衡的條件下有效地偵測淋巴球細胞。為了解決以上提及的問題，本文提出了一個主動學習框架來減少總標記工作量，並通過下採樣多數類別族群和過採樣少數類別族群，同時解決極端類別不平衡問題。實驗結果顯示，主動學習可以提供極度資料不平衡的數據於分割任務中有效的解。該方法的貢獻有以下三個，（1）我們提出一個主動學習框架，其透過下採樣多數類別族群和過採樣少數類別族群來解決極度資料不平衡問題。（2）並且於肝臟組織切片影像中以較少量的訓練資料得到高性能的結果。最後，（3）提供淋巴細胞的量化分析以進行更客觀的診斷。

Medical images segmentation is a fundamental challenge in medical image analysis. A major concern in the application of biomedical images in deep learning is insufficient number of annotated samples. Since the annotation process requires specialty-oriented knowledge and there are often too many instances in images (e.g. cells), this can incur a great deal of annotation effort and cost. Another concern is class imbalance problem, which is a critical obstacle commonly occurred in biomedical images. Considering the application of lymphocyte detection, an important lymphocyte subpopulation is extremely fewer than other cells, which would make training more biased toward the majority class. However, traditional labeling strategies, such as active learning, are ineffective in finding enough minority samples to train. Hence, this study deploys a low-cost method for manual annotation for efficiently lymphocyte detection in domains exhibiting extreme class imbalance. To address these problems, this paper proposed an active learning framework to reduce the total labeled workload while solving the extreme class imbalance problem by both under-sampling majority class and over-sampling minority class. Experimental results show that the proposed method can achieve annotation-effective solution in extremely imbalanced class segmentation. The contribution of the proposed method has three-fold, (1) we proposed an AL framework for solving the extreme class imbalance problem by both under-sampling majority class and over-sampling minority class. (2) the proposed framework achieves good performance for lymphocyte detection in histopathological image with fewer labeled samples. Finally, (3) quantitative analysis of lymphocytes is provided for more objective diagnosis.

摘要 I
Abstract III
Table of Content VI
List of Tables VIII
List of Figures VIII
Chapter 1 Introduction 1
1.1 Motivation 1
1.2 Purpose 3
1.3 Challenge of lymphocyte detection in histopathology image 7
1.4 Active learning 11
1.5 Proposed method 13
Chapter 2 Related Works 15
2.1 CNN methods for medical image segmentation 15
2.2 Imbalanced data 16
2.2.1 Resampling 16
2.2.2 Minority object penalty loss 17
2.3 Lessen the burden of manual annotation 17
2.3.1 Transfer learning for medical image 17
2.3.2 Active learning methods 18
Chapter 3 Proposed Method 21
3.1 Histopathology Images Acquisition 22
3.2 Annotation-effective Active learning 24
3.3 Active candidate selection 25
3.3.1 Model Uncertainty Estimation 25
3.3.2 Randomness 29
3.3.3 Minority sampling 29
3.3.4 Minority data-augmentation 30
3.4 Minority-Sensitive Dice Loss 31
3.5 Continuously fine-tuning 34
3.6 Implementation Details 35
3.7 CNN Models 36
Chapter 4 Experimental Results and Discussions 39
4.1 H＆E Stained Liver Tissue Dataset 39
4.2 Evaluation Criterion 42
4.3 Make Balance vs. Minority Sampling 43
4.4 Performance of AL_MB_MS 44
4.5 Performance of High Skew Ratio in Minority samples 47
4.7 Result of Cell Detection 49
4.8 Quantified result of lymphocyte detection 53
4.9 Platform 53
Chapter 5 Conclusion and future work 55
References 56

[1]Zhu, R. X., Seto, W. K., Lai, C. L., ＆ Yuen, M. F. (2016). Epidemiology of hepatocellular carcinoma in the Asia-Pacific region. Gut and liver, 10(3), 332.
[2]Fridman, W. H., Pages, F., Sautes-Fridman, C., ＆ Galon, J. (2012). The immune contexture in human tumours: impact on clinical outcome. Nature Reviews Cancer, 12(4), 298.
[3]Ishak, K., Baptista, A., Bianchi, L., Callea, F., De Groote, J., Gudat, F., ... ＆ Phillips, M. J. (1995). Histological grading and staging of chronic hepatitis. Journal of hepatology, 22(6), 696-699.
[4]https://kknews.cc/zh-tw/health/pbk2xp.html
[5]https://www.britannica.com/science/liver
[6]Lai, C. L., Ratziu, V., Yuen, M. F., ＆ Poynard, T. (2003). Viral hepatitis B. The Lancet, 362(9401), 2089-2094.
[7]Zhou, Z., Shin, J., Zhang, L., Gurudu, S., Gotway, M., ＆ Liang, J. (2017). Fine-tuning convolutional neural networks for biomedical image analysis: actively and incrementally. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 7340-7351).
[8]Wang, K., Zhang, D., Li, Y., Zhang, R., ＆ Lin, L. (2016). Cost-effective active learning for deep image classification. IEEE Transactions on Circuits and Systems for Video Technology, 27(12), 2591-2600.
[9]Yang, L., Zhang, Y., Chen, J., Zhang, S., ＆ Chen, D. Z. (2017, September). Suggestive annotation: A deep active learning framework for biomedical image segmentation. In International conference on medical image computing and computer-assisted intervention (pp. 399-407). Springer, Cham.
[10]Gorriz, M., Carlier, A., Faure, E., ＆ Giro-i-Nieto, X. (2017). Cost-effective active learning for melanoma segmentation. arXiv preprint arXiv:1711.09168.
[11]Ozdemir, F., Peng, Z., Tanner, C., Fuernstahl, P., ＆ Goksel, O. (2018). Active learning for segmentation by optimizing content information for maximal entropy. In Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support (pp. 183-191). Springer, Cham.
[12]Mackowiak, R., Lenz, P., Ghori, O., Diego, F., Lange, O., ＆ Rother, C. (2018). Cereals-cost-effective region-based active learning for semantic segmentation. arXiv preprint arXiv:1810.09726.
[13]Modified HAI Scoring System (Ishak et al., 1995)
[14]He, K., Girshick, R., ＆ Dollár, P. (2018). Rethinking imagenet pre-training. arXiv preprint arXiv:1811.08883.
[15]Cao, H., Bernard, S., Heutte, L., ＆ Sabourin, R. (2018, June). Improve the performance of transfer learning without fine-tuning using dissimilarity-based multi-view learning for breast cancer histology images. In International Conference Image Analysis and Recognition (pp. 779-787). Springer, Cham.
[16]Yosinski, J., Clune, J., Bengio, Y., ＆ Lipson, H. (2014). How transferable are features in deep neural networks? In Advances in neural information processing systems (pp. 3320-3328).
[17]Milletari, F., Navab, N., ＆ Ahmadi, S. A. (2016, October). V-net: Fully convolutional neural networks for volumetric medical image segmentation. In 2016 Fourth International Conference on 3D Vision (3DV) (pp. 565-571). IEEE.
[18]Oktay, O., Schlemper, J., Folgoc, L. L., Lee, M., Heinrich, M., Misawa, K., ... ＆ Glocker, B. (2018). Attention U-Net: learning where to look for the pancreas. arXiv preprint arXiv:1804.03999.
[19]Ronneberger, O., Fischer, P., ＆ Brox, T. (2015, October). U-net: Convolutional networks for biomedical image segmentation. In International Conference on Medical image computing and computer-assisted intervention (pp. 234-241). Springer, Cham.
[20]Hu, B., Yang, X. R., Xu, Y., Sun, Y. F., Sun, C., Guo, W., ... ＆ Fan, J. (2014). Systemic immune-inflammation index predicts prognosis of patients after curative resection for hepatocellular carcinoma. Clinical Cancer Research, 20(23), 6212-6222.
[21]Kane, R. C., Farrell, A. T., Madabushi, R., Booth, B., Chattopadhyay, S., Sridhara, R., ... ＆ Pazdur, R. (2009). Sorafenib for the treatment of unresectable hepatocellular carcinoma. The oncologist, 14(1), 95-100.
[22]Kobayashi, N., Usui, S., Kikuchi, S., Goto, Y., Sakai, M., Onizuka, M., ＆ Sato, Y. (2012). Preoperative lymphocyte count is an independent prognostic factor in node-negative non-small cell lung cancer. Lung cancer, 75(2), 223-227.
[23]Ertekin, S., Huang, J., Bottou, L., ＆ Giles, L. (2007, November). Learning on the border: active learning in imbalanced data classification. In Proceedings of the sixteenth ACM conference on Conference on information and knowledge management (pp. 127-136). ACM.
[24]Gal, Y., ＆ Ghahramani, Z. (2016, June). Dropout as a bayesian approximation: Representing model uncertainty in deep learning. In international conference on machine learning (pp. 1050-1059).
[25]Kendall, A., Badrinarayanan, V., ＆ Cipolla, R. (2015). Bayesian segnet: Model uncertainty in deep convolutional encoder-decoder architectures for scene understanding. arXiv preprint arXiv:1511.02680.
[26]Kubat, M., ＆ Matwin, S. (1997, July). Addressing the curse of imbalanced training sets: one-sided selection. In Icml (Vol. 97, pp. 179-186).
[27]Chawla, N. V., Bowyer, K. W., Hall, L. O., ＆ Kegelmeyer, W. P. (2002). SMOTE: synthetic minority over-sampling technique. Journal of artificial intelligence research, 16, 321-357.
[28]Lin, T. Y., Goyal, P., Girshick, R., He, K., ＆ Dollár, P. (2017). Focal loss for dense object detection. In Proceedings of the IEEE international conference on computer vision (pp. 2980-2988).
[29]Fu, C., Qu, W., ＆ Yang, Y. (2013). Actively learning from mistakes in class imbalance problems. IFAC Proceedings Volumes, 46(13), 341-346.
[30]https://www.pathpedia.com/education/eatlas/histopathology/general_pathology/chronic_inflammation_-_plasma_cells.aspx