臺灣博碩士論文加值系統

以往影像辨識已應用於許多地方，相較於醫學影像辨識方面卻是比較少見，而近年來由於深度學習於影像辨識領域中有很大的進步，因此越來越多學者關注到這方面的研究上。在本論文中，我提出一個改良式卷積神經網路應用於醫學影像之辨識，藉由從Kaggle平台上得到的開放胸腔X-Ray影像圖進行肺炎預測辨識之應用，首先我們取得的資料數量較為少量，且分類後資料筆數不平衡，對於神經網路訓練，資料量過少和資料不平衡，很容易導致訓練出之模型辨識率極差。因此本論在第一部分進行資料前處理，利用資料增強的方式弭補資料量過少與資料不平衡之問題；第二部分則使用各知名卷積神經網路和本論文提出之IVGG13訓練於未做資料增強的原始資料，並評估訓練後各模型之辨識率及其他評估值；最後於第三部分為使用各知名卷積神經網路和IVGG13訓練於已做資料增強後的資料集，訓練後評估個模型之辨識率與各項評估標準。經過上述步驟後，於研究結果分析與比較中，可以透過實驗結果中的評估結果證明本論文提出之IVGG13於醫學影像辨識優於其他知名卷積神經網路，且使用資料增強能夠有效提升模型之準確率。

In the past, image recognition has been applied to many places, but it is relatively rare compared with medical image recognition. In recent years, due to the great progress of deep learning in the field of image recognition, more and more scholars are paying attention to this research. In this paper, I propose an improved convolutional neural network for medical image recognition. The application of pneumonia predictive identification by open source chest X-ray images obtained from the Kaggle platform. First of all, the amount of data we obtained is relatively small, and the number of data after classification is not balanced. For neural network training, the amount of data is too small and the data are unbalanced, which can easily lead to a poor recognition rate of the model. Therefore, in the first part, I use the data enhancement method to supplement the problem of data too little and data imbalance. In the second part, I use the well-known convolutional neural networks and IVGG13 proposed in this paper to train the originals without data enhancement, and evaluate the identification rate and other evaluation values of each model after training. Finally, I use the well-known convolutional neural networks and IVGG13 to train the datasets that have been enhanced, and evaluate the identification rate and other evaluation values of each model after training. After the above steps, in the analysis and comparison of the research results, the evaluation results in the experimental results can be used to prove that the IVGG13 proposed in this paper is superior to other well-known convolutional neural networks in medical image recognition, and the use of data enhancement can effectively improve the accuracy of models.

摘要
Abstract
誌謝
目錄
表目錄
圖目錄
一、緒論
1.1 研究背景
1.2 研究動機與目的
1.3 論文架構
二、文獻回顧
2.1 醫學影像
2.1.1 X-Ray
2.1.2 Computed Tomography
2.1.3 Magnetic Resonance Imaging
2.1.4 Positron Emission Tomography
2.1.5 Single Photon Emission Computed Tomography
2.1.6 Medical Ultrasound
2.2 機器學習
2.2.1 學習模式
2.3 類神經網路
2.3.1 激活函數
2.4 卷積神經網路
2.4.1 卷積網路架構
2.5 深度學習
2.5.1 深度學習模型
2.6 Keras
2.7 TensorFlow
2.8 Kaggle
三、研究方法
3.1訓練資料集
3.2資料增強(Data Augmentation)
3.3 IVGG13
四、 研究結果分析與比較
4.1 比較之卷積神經網路
4.1.1 LeNet
4.1.2 AlexNet
4.1.3 VGG16
4.2 未做前處理前之評估結果
4.3 做完前處理後之評估結果
4.4 分析結果發現VGG16之問題
五、 結論與未來展望
5.1 結論
5.2 未來展望
參考文獻

[1]Hewitt, C., Bishop, P., & Steiger, R. (1973, August). A universal modular actor formalism for artificial intelligence. In Proceedings of the 3rd international joint conference on Artificial intelligence(pp. 235-245).
[2]Bishop, C. M. (2006). Pattern recognition and machine learning. springer.
[3]Y. LeCun, Y. Bengio, and G. Hinton, ‘‘ Deep learning,’’ Nature, 521(7553), 436-444, 2015.
[4]McCarthy, J., Minsky, M. L., Rochester, N., & Shannon, C. E. (2006). A proposal for the dartmouth summer research project on artificial intelligence, august 31, 1955. AI magazine, 27(4), 12-12.
[5]Silver, D. et al.,“Mastering the Game of Go with Deep Neural Networks and Tree Search”, Nature, vol. 529, pp. 484-489, 2016.
[6]A. Criminisi, ‘‘Machine learning for medical images analysis,’’ 2016.
[7]C. Affonso, A. L. D. Rossi, F. H. A. Vieira, and A. C. P. de Leon Ferreira de Carvalho, ‘‘Deep learning for biological image classification,’’ Expert Systems with Applications, 85, 114-122, 2017.
[8]D. Norris, ‘‘Shortlist: A connectionist model of continuous speech recognition,’’ Cognition, 52(3), 189-234, 1994.
[9]G. G. Chowdhury, ‘‘Natural language processing,’’ Annual review of information science and technology, 37(1), 51-89, 2003.
[10]Gardner, M. W., & Dorling, S. R. (1998). Artificial neural networks (the multilayer perceptron)—a review of applications in the atmospheric sciences. Atmospheric environment, 32(14-15), 2627-2636.
[11]Hinton, Geoffrey, et al. "Deep neural networks for acoustic modeling in speech recognition." IEEE Signal processing magazine 29 (2012).
[12]Mikolov, T., Karafiát, M., Burget, L., Černocký, J., & Khudanpur, S. (2010). Recurrent neural network based language model. In Eleventh annual conference of the international speech communication association.
[13]Russakovsky, Olga, et al. "Imagenet large scale visual recognition challenge." International Journal of Computer Vision 115.3 (2015): 211-252.
[14]Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. In Advances in neural information processing systems (pp. 1097-1105).
[15]LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278-2324.
[16]Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., ... & Rabinovich, A. (2015). Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 1-9).
[17]Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.
[18]Röntgen, W. C. (1898). Ueber eine neue Art von Strahlen. Annalen der Physik, 300(1), 1-11.
[19]Hounsfield, G. N. (1995). Computerized transverse axial scanning (tomography): Part I. Description of system. 1973.
[20]Ogawa, S., Lee, T. M., Kay, A. R., & Tank, D. W. (1990). Brain magnetic resonance imaging with contrast dependent on blood oxygenation. proceedings of the National Academy of Sciences, 87(24), 9868-9872.
[21]Bailey, D. L., Maisey, M. N., Townsend, D. W., & Valk, P. E. (2005). Positron emission tomography. London: Springer.
[22]Wagner, A., Mahrholdt, H., Holly, T. A., Elliott, M. D., Regenfus, M., Parker, M., ... & Judd, R. M. (2003). Contrast-enhanced MRI and routine single photon emission computed tomography (SPECT) perfusion imaging for detection of subendocardial myocardial infarcts: an imaging study. The Lancet, 361(9355), 374-379.
[23]Hoff, L. (2001). Acoustic characterization of contrast agents for medical ultrasound imaging. Springer Science & Business Media.
[24]Russell, Stuart J., and Peter Norvig. Artificial intelligence: a modern approach. Malaysia; Pearson Education Limited, 2016.
[25]Hastie, Trevor, Robert Tibshirani, and Jerome Friedman. "Unsupervised learning." The elements of statistical learning. Springer, New York, NY, 2009. 485-585.
[26]Kaelbling, L. P., Littman, M. L., & Moore, A. W. (1996). Reinforcement learning: A survey. Journal of artificial intelligence research, 4, 237-285.
[27]林大貴，2017，TensorFlow + Keras 深度學習人工智慧實務應用，初版，博碩文化股份有限公司，新北市
[28]Hartigan, J. A., & Wong, M. A. (1979). Algorithm AS 136: A k-means clustering algorithm. Journal of the Royal Statistical Society. Series C (Applied Statistics), 28(1), 100-108.
[29]Agrawal, R., & Srikant, R. (1994, September). Fast algorithms for mining association rules. In Proc. 20th int. conf. very large data bases, VLDB (Vol. 1215, pp. 487-499).
[30]Danielsson, Per-Erik. "Euclidean distance mapping." Computer Graphics and image processing 14.3 (1980): 227-248.
[31]Schultz, W., Dayan, P., & Montague, P. R. (1997). A neural substrate of prediction and reward. Science, 275(5306), 1593-1599.
[32]N. Gupta, ‘‘Artificial neural network,’’ Network and Complex Systems, 3(1), 24-28, 2013.
[33]W. Cao, X. Wang, Z. Ming, and J. Gao, ‘‘A review on neural networks with random weights,’’ Neurocomputing, 2017.
[34]高揚，2018，白話深度學習與TensorFlow，初版，碁峰資訊股份有限公司，台北市
[35]斎藤康毅，2017，Deep Learning :用python進行深度學習的基礎理論實作，吳嘉芳，初版，碁峰資訊股份有限公司，台北市
[36]Hubel, D. H., & Wiesel, T. N. (1962). Receptive fields, binocular interaction and functional architecture in the cat's visual cortex. The Journal of physiology, 160(1), 106-154.
[37]Fukushima, K. (1980). Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position. Biological cybernetics, 36(4), 193-202.
[38]Deng, Jia, et al. "Imagenet: A large-scale hierarchical image database." Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on. Ieee, 2009. (http://dx.doi.org/10.1109/CVPR.2009.5206848)
[39]黃文堅、唐源，2017，實戰 TensorFlow Google 深度學習，蔡勝文，初版，碁峰資訊股份有限公司，台北市
[40]Nick Locascio，2018，Deep Learning深度學習基礎｜設計下一代人工智慧演算法，藍子軒，初版，碁峰資訊股份有限公司，台北市
[41]D. Kermany, K. Zhang, and M. Goldbaum, “Labeled optical coherence tomography (OCT) and Chest X-Ray images for classification,” Mendeley Data, v2, 2018, Available: http://dx.doi.org/10.17632/rscbjbr9sj.2
[42]Jabbar, H., & Khan, D. R. Z. (2015). Methods to avoid over-fitting and under-fitting in supervised machine learning (comparative study). Computer Science, Communication and Instrumentation Devices.