研究生: |
藍友廷 You-Ting Lan |
---|---|
論文名稱: |
卷積神經網路於DNA影像自我交叉辨識之應用 Application of Convolutional Neural Network in DNA Image Self-Crossing Recognition |
指導教授: |
張以全
I-Tsyuen Chang |
口試委員: |
張以全
I-Tsyuen Chang 林顯易 Hsien-I Lin 藍振洋 Chen-yang Lan |
學位類別: |
碩士 Master |
系所名稱: |
工程學院 - 機械工程系 Department of Mechanical Engineering |
論文出版年: | 2020 |
畢業學年度: | 108 |
語文別: | 中文 |
論文頁數: | 85 |
中文關鍵詞: | DNA影像 、遷移學習 、自我交叉 、卷積神經網路 、影像辨識 |
外文關鍵詞: | DNA image, Transfer learning, Self-crossing, CNN, Image classification |
相關次數: | 點閱:316 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 |
本論文於DNA樣本分析領域中,針對使用原子力顯微鏡(Atomic Force Microscope)所獲取的DNA樣本數位影像為研究對象,以深度學習的技術發展出一個自動化辨識DNA交叉與非交叉樣本的新方法。本論文透過DNA長鏈模型模擬DNA結構,再經過數位化處理得到類似原子力顯微鏡的影像,藉此建立過去無法建構的DNA影像資料庫,解決DNA樣本難以取得的問題。根據所模擬出的影像選用適合的經典卷積神經網路,再基於遷移學習(Transfer Learning)技術建構出針對此任務目的的卷積神經網路(Convolutional Neural Network)模型,以大量交叉與非交叉影像對其進行訓練,而得到比過去以Rule Based方法更佳的準確率。在本研究中也參入影像雜訊對神經網路模型進行訓練,並探討以深度學習方法與過去辨識方法兩者的表現。本論文提出上述方法可以降低對樣本影像前處理的依賴性,改善遇到無法順利前處理或是前處理過後依然無法以傳統影像處理方式順利辨識交叉特徵的問題,增加所能應付的影像類型,並且提高辨識的準確率。
In the field of analyzing DNA samples, this thesis takes the digital images of DNA samples, obtained by Atomic Force Microscope (AFM), as the research object, and develops a novel method to automatically classify whether DNA is crossing or not by deep learning technology. In this thesis, the DNA structure is simulated through the DNA long-chain model. After that, through digitization, an image, similar to one obtained by AFM, can be acquired. A DNA image database, which cannot be built in the past, is established by the steps above, and it solves the problem that the DNA samples are difficult to obtain. An appropriate, classical Convolutional Neural Network (CNN) is selected according to the simulated images, and based on the technique of Transfer Learning, the CNN model for this task is established. With the training of a large number of images of crossing and non-crossing, a better accuracy can be achieved compared to the Rule-Based method. In this research, images with noise are also applied to train the neural network model, and the performance of deep learning method and the Rule-Based method, which is applied in the past, will be discussed. With the deep learning method that proposed in this thesis, it can reduce the reliance of the pre-processing of the sample images. In other words, it improves the problems, such as applying the pre-processing unsuccessfully, or still cannot classify the features of crossing even after the pre-processing. Furthermore, this method not only increases the types of images that can be handled, but also achieves a higher accuracy of classification.
[1] D. Y. Abramovitch, S. B. Andersson, L. Y. Pao, and G. Schitter, “A tutorial on the mechanisms, dynamics, and control of atomic force microscopes,” in American Control Conference, 2007. ACC’07, pp. 3488–3502, IEEE, 2007.
[2] “https://en.wikipedia.org/wiki/dna,” 2018.
[3] K. Drlica et al., “Understanding dna and gene cloning: a guide for the curious.,”
Understanding DNA and Gene Cloning: a Guide for the Curious., no. ed. 2, 1992.
[4] Y. LeCun, “https://www.slideshare.net/yandex/yann-le-cun.”
[5] J. Bednar, P. Furrer, V. Katritch, A. Stasiak, J. Dubochet, and A. Stasiak, “Deter- mination of DNA persistence length by cryoelectron microscopy. separation of the static and dynamic contributions to the apparent persistence length of A,” Journal of Molecular Biology, vol. 254, no. 4, pp. 579–594, 1995.
[6] C.Rivetti,M.Guthold,andC.Bustamante,“ScanningforcemicroscopyofDNAde- posited onto mica: Equilibration versus kinetic trapping studied by statistical poly- mer chain analysis,” Journal of Molecular Biology, vol. 264, no. 5, pp. 919–932, 1996.
[7] H. Wang and J. N. Milstein, “Simulation assisted analysis of the intrinsic stiffness for short DNA molecules imaged with scanning atomic force microscopy,” PloS one, vol. 10, no. 11, p. e0142277, 2015.
[8] H. Freeman, “On the encoding of arbitrary geometric configurations,” IRE Transac- tions on Electronic Computers, no. 2, pp. 260–268, 1961.
[9] H. Freeman, “Computer processing of line-drawing images,” ACM Computing Sur- veys (CSUR), vol. 6, no. 1, pp. 57–97, 1974.
[10] T. Spisz, Y. Fang, R. Reeves, C. Seymour, I. Bankman, and J. Hoh, “Automated siz- ing of DNA fragments in atomic force microscope images,” Medical and Biological Engineering and Computing, vol. 36, no. 6, pp. 667–672, 1998.
[11] C. Rivetti, “A simple and optimized length estimator for digitized DNA contours,” Cytometry Part A, vol. 75, no. 10, pp. 854–861, 2009.
[12] Z. Kulpa, “Area and perimeter measurement of blobs in discrete binary pictures,” Computer Graphics and Image Processing, vol. 6, no. 5, pp. 434–451, 1977.
[13] E.Ficarra,L.Benini,E.Macii,andG.Zuccheri,“AutomatedDNAfragmentsrecog- nition and sizing through AFM image processing,” IEEE Transactions on Informa- tion Technology in Biomedicine, vol. 9, no. 4, pp. 508–517, 2005.
[14] A. Sundstrom, S. Cirrone, S. Paxia, C. Hsueh, R. Kjolby, J. K. Gimzewski, J. Reed, and B. Mishra, “Image analysis and length estimation of biomolecules using AFM,” IEEE Transactions on Information Technology in Biomedicine, vol. 16, no. 6, pp. 1200–1207, 2012.
[15] P.-I.ChangandM.-C.Hsaio,“Resolution-freeaccurateDNAcontourlengthestima- tion from atomic force microscopy images,” Scanning, 2019.
[16] M.-C. Hsaio, “DNA contour length estimator utilizing shape number from AFM imaging,” National Taiwan Universiry of science and technology, 2016.
[17] C. Rivetti, C. Walker, and C. Bustamante, “Polymer chain statistics and conforma- tional analysis of DNA molecules with bends or sections of different flexibility,” Journal of molecular biology, vol. 280, no. 1, pp. 41–59, 1998.
[18] Z.-X.Dai,“Accurateestimationfordigitalcurvetangentvectorofbiologicalsamples from AFM systems,” National Taiwan Universiry of science and technology, 2018.
[19] I.-J.Yu,“EstimationofDNApersistencelengthwithatomicforcemicroscopyimag- ing,” National Taiwan Universiry of science and technology, 2019.
[20] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet classification with deep convolutional neural networks,” in Advances in Neural Information Processing Sys- tems 25 (F. Pereira, C. J. C. Burges, L. Bottou, and K. Q. Weinberger, eds.), pp. 1097– 1105, Curran Associates, Inc., 2012.
[21] K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” 2014.
[22] Z. Wang, W. Zheng, C. Song, Z. Zhang, J. Lian, S. Yue, and S. Ji, “Air quality mea- surement based on double-channel convolutional neural network ensemble learn- ing,” IEEE Access, vol. 7, pp. 145067–145081, 2019.
[23] Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner, “Gradient-based learning applied to document recognition,” Proceedings of the IEEE, vol. 86, no. 11, pp. 2278–2324, 1998.
[24] C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V. Van- houcke, and A. Rabinovich, “Going deeper with convolutions,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2015.
[25] A.Papoulis,Probability,randomvariables,andstochasticprocesses.McGraw-Hill, 1965.
[26] K. Pearson, “The problem of the random walk,” Nature, vol. 72, no. 1867, p. 342, 1905.
[27] L. Lam, S.-W. Lee, and C. Y. Suen, “Thinning methodologies-a comprehensive sur- vey,” IEEE Transactions on pattern analysis and machine intelligence, vol. 14, no. 9, pp. 869–885, 1992.
[28] “https://www.mathworks.com/matlabcentral/fileexchange/13351-fast-and-robust- self-intersections.”
[29] S. B. Gray, “Local properties of binary images in two dimensions,” IEEE Transac- tions on Computers, vol. 100, no. 5, pp. 551–561, 1971.
[30] W. K. Pratt, “Digital image processing: PIKS scientific inside/William K. Pratt.– NY,” 2007.