生物學者在研究老化的機制時常會使用秀麗隱桿線蟲做為樣本分析其老化特徵以決定其生理日齡，因此如何精確評量秀麗隱桿線蟲的生理年齡是很重要且基本的課題。本論文主要設計針對秀麗隱桿線蟲的日齡預估模型。針對陽明交通大學許翱麟教授所提供之線蟲顯微影像，透過建立深度神經網路的方式取代以往生物研究人員依賴肉眼及經驗的方式做線蟲的日齡判讀，以協助研究人員進行評估。
　　本論文使用幾種過往經典的輕量型神經網路模型作為基底，包含VGG16、ResNet50、InceptionResNet、Xception及EfficientNet-B1，進行秀麗隱桿線蟲日齡迴歸分析，經過平均絕對誤差、模型參數量的評比考量後選取了EfficientNet-B1[25]作為判讀線蟲日齡的深度神經網路模型。而實驗觀察得知線蟲在日齡超過三天並進入成蟲期後，其體長的成長速度會趨於飽和[21]，因此需要針對線蟲蟲體的內臟空洞程度，使用注意力機制觀察其紋理上的變化，減少日齡大於三天的成蟲日齡的錯判情形。本論文嘗試了三種注意力機制，包括SENet (Squeeze-and Excitation Network)、Spatial Attention、與CBAM (Convolutional Block Attention Module)，並經實驗找出最佳的模型與注意力機制組合，將誤差為一天內的準確率提升至72.47%，誤差為兩天內的準確率提升至92.42%。

　　Biologists often use Caenorhabditis elegans as a sample to analyze the aging characteristics to determine its physiological age when studying the mechanism of aging. Therefore, how to accurately evaluate the physiological age of C. elegans is an important and basic topic. This thesis mainly designs the age prediction model for C. elegans. A deep neural network with attention mechanism is built using the microscopic images of nematodes provided by Professor Hsu, Ao-Lin of Yang Ming Chiao Tung University. The established model replaces the previous method of biological researchers which relies on the naked eye and experience to interpret the age of the nematode, and thus assists the researchers in the age evaluation process.
This thesis uses several classic lightweight neural network models as the base, including VGG16, ResNet50, InceptionResNet, Xception and EfficientNet-B1, to carry out the regression analysis of the age of C. elegans. After consideration of the mean absolute error and the evaluation of the numbers of model parameters, EfficientNet-B1[23] was selected as the deep neural network model for interpreting the age of nematodes. The experimental observation shows that when the nematode is older than three days and enters the adult stage, the growth rate of its body length will tend to be saturated [21], so it is necessary to use the attention mechanism to observe the texture of the visceral cavity of the nematode body. The changes in the larvae reduce the misjudgment of the age of adults older than three days. This paper tries three attention mechanisms, including SENet (Squeeze-and Excitation Network), Spatial Attention, and CBAM (Convolutional Block Attention Module), and finds the best combination of model and attention mechanism through experiments, and the accuracy of the prediction is increased to 72.47% within one day error, and the accuracy of the prediction is increased to 92.42% within two days error.

封面 1
摘要 I
Abstract II
致謝 IV
目錄 V
圖目錄 VIII
表目錄 XI
第一章 緒論 1
1.1 研究背景 1
1.2 研究動機與文獻回顧 2
1.3 論文架構 4
第二章 機器學習與神經網路 5
2.1 前言 5
2.2 機器學習（Machine Learning) 5
2.3 監督式學習（Supervised Learning） 8
2.3.1 迴歸分析（Regression Analysis） 9
2.3.2 分類問題（Classification） 14
2.4 分類器（Classifier） 15
2.4.1 邏輯回歸（Logistic Regression）[38][39] 15
2.4.2 貝氏分類器（Bayesian Classifier）[40] 18
2.4.3 支持向量機（support vector machines）[41][42] 20
2.4.4 最近鄰居分類器（K-Nearest-Neighbor Classifier）[43] 21
2.4.5 決策樹（Decision Tree）[44] 22
第三章 神經網路模型 24
3.1 前言 24
3.2 VGG16（Visual Geometry Group-16）[45] 24
3.3 ResNet50（Residual neural network 50）[46] 26
3.4 Inception-ResNet[48] 28
3.5 Xception（Extreme version of Inception）[50] 31
3.6 EfficientNet[51] 33
3.7 注意力機制（Attention mechanism） 38
3.7.1 SENet（Squeeze-and-Excitation Networks）[52] 39
3.7.2 CBAM（Convolutional Block Attention Module）[54] 41
第四章 資料集與預處理 44
4.1 前言 44
4.2 資料集 44
4.3 資料前處理 47
4.3.1 資料分割 48
4.3.2 資料擴增 48
第五章 實驗方法與結果 52
5.1 前言 52
5.2 模型預測類型 52
5.3 實驗環境設置 52
5.4 實驗步驟 53
5.5 模型選擇與比較 54
5.6 使用EfficientNetB1結合CBAM制架構實驗結果 56
5.7 EfficientNet-B1結合其餘注意力機制實驗結果 61
第六章 結論與未來展望 65
6.1 結論 65
6.2 未來展望 65
參考文獻 67

[1] Rachel A. Ankeny and Sabina Leonelli, “What’s so special about model organisms?,” Studies in History and Philosophy of Science 42 (2), 2011, pp. 313-323. doi:10.1016/j.shpsa.2010.11.039
[2] Zhang S., Zhou T., Wang G., and Li Z., “Caenorhabditis elegans as a Useful Model for Studying Aging Mutations,” Front Endocrinol (Lausanne)., 2020. doi: 10.3389/fendo.2020.554994.
[3] Kaletta T. and Hengartner M.O., “Finding function in novel targets: C. elegans as a model organism,” Nat Rev Drug Discov. 5(5), 2006, pp. 387-398. doi: 10.1038/nrd2031.
[4] Lord C. E. and Gunawardena A. H., “Programmed cell death in C. elegans, mammals and plants,” Eur J Cell Biol. 91(8), 2012. doi: 10.1016/j.ejcb.2012.02.002.
[5] Conradt B., Wu Y.C., and Xue D., “Programmed Cell Death During Caenorhabditis elegans Development,” Genetics. 203(4), 2016. doi: 10.1534/genetics.115.186247.
[6] Lee Gee-Yoon, Sohn J., and Lee S.V., “Combinatorial Approach Using Caenorhabditis elegans and Mammalian Systems for Aging Research,” Mol Cells 44(7), 2021, pp. 425-432. doi:10.14348/molcells.2021.0080.
[7] Gruber J., Chen C.B., Fong S., Ng L.F., Teo E., and Halliwell B., “Caenorhabditis elegans: What We Can and Cannot Learn from Aging Worms,” Antioxid Redox Signal 20(3), 2015, doi: 10.1089/ars.2014.6210.
[8] Hekimi S., Lakowski B., Barnes TM, and Ewbank J.J., “Molecular genetics of life span in C. elegans: how much does it teach us?,” Trends Genet. 14(1), 1998, pp. 14-20. doi: 10.1016/S0168-9525(97)01299-7.
[9] Muschiol D., Schroeder F., and Traunspurger W., “Life cycle and population growth rate of Caenorhabditis elegans studied by a new method,” BMC Ecol 9(14), 2009. doi: 10.1186/1472-6785-9-14.
[10] Tissenbaum H.A., “Using C. elegans for aging research,” Invertebr Reprod Dev. 59, pp. 59-63. doi: 10.1080/07924259.2014.940470.
[11] Woodruff, G.C., Johnson, E., and Phillips, P.C., “A large close relative of C. elegans is slow-developing but not long-lived,” BMC Evol Biol 19 (74), 2019. doi: 10.1186/s12862-019-1388-1.
[12] Zhang W.B., Sinha D.B., Pittman W.E., Hvatum E., Stroustrup N., and Pincus Z., “Extended Twilight among Isogenic C. elegans Causes a Disproportionate Scaling between Lifespan and Health,” Cell Syst. 3(4), 2016, pp. 333-345. doi: 10.1016/j.cels.2016.09.003.
[13] Feng Z. et al., “An imaging system for standardized quantitative analysis of C. elegans behavior,” BMC Bioinformatics 5(115), 2004, doi: 10.1186/1471-2105-5-115.
[14] Porta-de-la-Riva M., Fontrodona L., Villanueva A., and Cerón J., “Basic Caenorhabditis elegans methods: synchronization and observation,” Journal of Visualized Experiments 64, 2012. doi: 10.3791/4019.
[15] Cheng Huang, Chengjie Xiong, and Kerry Kornfeld, “Measurements of age-related changes of physiological processes that predict lifespan of Caenorhabditis elegans,” The Proceedings of the National Academy of Sciences 101(21), 2004, pp. 8084-8089. doi: 10.1073/pnas.0400848101.
[16] Ao-Lin Hsu, Zhaoyang Feng, Meng-Yin Hsieh, and X. Z. Shawn Xu, “Identification by machine vision of the rate of motor activity decline as a lifespan predictor in C. elegans,” Neurobiology of Aging 30(9), 2009, pp. 1498-1503, doi: 10.1016/j.neurobiolaging.2007.12.007.
[17] Liu J., Zhang B., Lei H., Feng Z., Liu J., Hsu A.L., and Xu X.Z., “Functional aging in the nervous system contributes to age-dependent motor activity decline in C. elegans,” Cell Metab. 18(3), 2013, pp. 392-402. doi: 10.1016/j.cmet.2013.08.007.
[18] Shamir L., Wolkow C.A., Goldberg I.G., “Quantitative measurement of aging using image texture entropy,” Bioinformatics. 25(23), 2009, doi: 10.1093/bioinformatics/btp571.
[19] Restif C. and Metaxas D., “Tracking the swimming motions of C. elegans worms with applications in aging studies,” Med Image Comput Comput Assist Interv. 11, 2008, pp. 35-42, doi: 10.1007/978-3-540-85988-8_5.
[20] Johnston J., Iser W.B., Chow D.K., Goldberg I.G., Wolkow C.A., “Quantitative image analysis reveals distinct structural transitions during aging in Caenorhabditis elegans tissues,” PLoS One. 3(7), 2008, doi: 10.1371/journal.pone.0002821.
[21] J. -L. Lin, W. -L. Kuo, Y. -H. Huang, T. -L. Jong, A. -L. Hsu, and W. -H. Hsu, "Using Convolutional Neural Networks to Measure the Physiological Age of Caenorhabditis elegans," IEEE/ACM Transactions on Computational Biology and Bioinformatics 18(6), 2021, pp. 2724-2732, doi: 10.1109/TCBB.2020.2971992.
[22] Lin, Jiunn-Liang, Combining Physiological Features and Convolution Neural Network to Measure the Physiological Age of Caenorhabditis elegans, PhD dissertation, Dept. of E.E., National Tsing Hua University, 2020.
[23] Kuo, Wei-Liang, Bone Age Assessment and C. elegans Age Prediction Using Deep Convolutional Neural Network, MS thesis, Dept. of E.E., National Tsing Hua University, 2018.
[24] Wei-Chen Lo, Caenorhabditis elegans Age Prediction and Visual Explanations Using Deep Convolutional Neural Networks, MS thesis, Dept. of E.E., National Tsing Hua University, 2020.
[25] Chien Te-Lun, Hybrid Integrated Gradient visual explaining Light-weight Deep convolutional Neural Network for Caenorhabditis elegans Age Prediction, MS thesis, Dept. of E.E., National Tsing Hua University, 2021.
[26] Huang Hsiao-Ni, Combining XRAI and Integrated Gradients, Smooth Guided Grad-CAM techniques for visual explaining a deep convolutional neural network-EfficientNet for Caenorhabditis elegans age prediction, MS thesis, Dept. of E.E., National Tsing Hua University, 2021.
[27] Pisharady P.K., Vadakkepat P., and Loh A.P., ”Attention Based Detection and Recognition of Hand Postures Against Complex Backgrounds,” International Journal of Computer Vision 101, pp. 403–419, 2013, doi: 10.1007/s11263-012-0560-5.
[28] G. Cao, Y. Zhou, D. Bollegala, and S. Luo, "Spatio-temporal Attention Model for Tactile Texture Recognition," RSJ International Conference on Intelligent Robots and Systems, 2020, pp. 9896-9902, doi: 10.1109/IROS45743.2020.9341333.
[29] Daniel P. Felker, Christine E. Robbins, and Mark A. McCormick, “Automation of C. elegans lifespan measurement,” Translational Medicine of Aging 4, 2020, pp. 1-10, doi: 10.1016/j.tma.2019.12.001.
[30] Collins J.J., Huang C., Hughes S., and Kornfeld K., “The measurement and analysis of age-related changes in Caenorhabditis elegans,” WormBook., 2008, pp. 1-21. doi: 10.1895/wormbook.1.137.1
[31] Nasteski, Vladimir, “An overview of the supervised machine learning methods,” Horizons 4, 2017, pp. 51-62, doi: 10.20544/HORIZONS.B.04.1.17.P05
[32] A. Singh, N. Thakur, and A. Sharma, "A review of supervised machine learning algorithms," 2016 3rd International Conference on Computing for Sustainable Global Development, 2016, pp. 1310-1315.
[33] H. U. Dike, Y. Zhou, K. K. Deveerasetty, and Q. Wu, "Unsupervised Learning Based On Artificial Neural Network: A Review," 2018 IEEE International Conference on Cyborg and Bionic Systems, 2018, pp. 322-327, doi: 10.1109/CBS.2018.8612259.
[34] Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin, ““Why Should I Trust You?”: Explaining the Predictions of Any Classifier,” International Conference on Knowledge Discovery and Data MiningAugust, 2016, pp. 1135–1144 doi: 10.1145/2939672.2939778.
[35] W. Qiang and Z. Zhongli, "Reinforcement learning model, algorithms and its application," 2011 International Conference on Mechatronic Science, Electric Engineering and Computer, 2011, pp. 1143-1146, doi: 10.1109/MEC.2011.6025669.
[36] D. Maulud and A. M. Abdulazeez, “A Review on Linear Regression Comprehensive in Machine Learning,” JASTT 1(4), 2020, pp. 140-147, doi: 10.38094/jastt1457.
[37] Harvey J. Motulsky and Lennart A. Ransnas, “Fitting curves to data using nonlinear regression: a practical and nonmathematical review,” The FASEB Journal 1(5), 1987, pp. 365-374, doi: 10.1096/fasebj.1.5.3315805.
[38] Chao-Ying Joanne Peng, Kuk Lida Lee and Gary M. Ingersoll, “An Introduction to Logistic Regression Analysis and Reporting,” The Journal of Educational Research 96(1), 2022, pp. 3-14, doi: 10.1080/00220670209598786.
[39] T. Haifley, "Linear logistic regression: an introduction," IEEE International Integrated Reliability Workshop Final Report, 2002., 2002, pp. 184-187, doi: 10.1109/IRWS.2002.1194264.
[40] Daniel Berrar, “Bayes' Theorem and Naive Bayes Classifier.” Encyclopedia of Bioinformatics and Computational Biology 1, 2019, pp. 403-412, doi: 10.1016/B978-0-12-809633-8.20473-1.
[41] Noble W., “What is a support vector machine?,” Nature Biotechnology 24, 2006, pp. 1565–1567, doi: 10.1038/nbt1206-1565.
[42] Durgesh Srivastava and Lekha Bhambhu, “Data classification using support vector machine,” Journal of Theoretical and Applied Information Technology 12(1), 2010, pp. 1-7.
[43] Zhang Zhongheng, “Introduction to machine learning: k-nearest neighbors,” Annals of translational medicine 4(11), 2016 218. doi:10.21037/atm.2016.03.37
[44] Adnan Mohsin Abdulazeez, “Classification Based on Decision Tree Algorithm for Machine Learning,” Journal of Applied Science and Technology Trends 2(1), 2021, pp. 20-28, doi: 10.38094/jastt20165.
[45] Karen Simonyan and Andrew Zisserman, “Very Deep Convolutional Networks for Large-Scale Image Recognition,”
[46] K. He, X. Zhang, S. Ren and J. Sun, "Deep Residual Learning for Image Recognition," 2016 IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 770-778, doi: 10.1109/CVPR.2016.90.
[47] He K., Zhang X., Ren S., and Sun J., “Identity Mappings in Deep Residual Networks,” European Conference on Computer Vision, 2016, pp. 630-645, doi: 10.1007/978-3-319-46493-0_38.
[48] Christian Szegedy, Sergey Ioffe, Vincent Vanhoucke, and Alex Alemi, “Inception-v4, inception-ResNet and the impact of residual connections on learning,” AAAI’ 17, 2017, pp. 4278-4284.
[49] Sanjeev Arora, Aditya Bhaskara, Rong Ge, Tengyu Ma, “Provable Bounds for Learning Some Deep Representations,” Proceedings of Machine Learning Research 32(1), 2014, pp. 584-592.
[50] F. Chollet, "Xception: Deep Learning with Depthwise Separable Convolutions," 2017 IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 1800-1807, doi: 10.1109/CVPR.2017.195.
[51] Mingxing Tan and Quoc V. Le, “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” PMLR 97, 2019, pp. 6105-6114.
[52] J. Hu, L. Shen and G. Sun, "Squeeze-and-Excitation Networks," 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018, pp. 7132-7141, doi: 10.1109/CVPR.2018.00745.
[53] L. Chen et al., "SCA-CNN: Spatial and Channel-Wise Attention in Convolutional Networks for Image Captioning," 2017 IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 6298-6306, doi: 10.1109/CVPR.2017.667.
[54] Sanghyun Woo, Jongchan Park, Joon-Young Lee, and In So Kweon, “CBAM: Convolutional Block Attention Module,” Proceedings of the European Conference on Computer Vision, 2018, pp. 3-19.
[55] Matthew D. Zeiler and Rob Fergus, “Visualizing and Understanding Convolutional Networks,” European Conference on Computer Vision, 2014, pp. 813-833. doi: 10.1007/978-3-319-10590-1_53.
[56] Kuo Wei-Liang, “Bone Age Assessment and C. elegans Age Prediction Using Deep Convolutional Neural Network,” 2018.
[57] Hildegard I.D. Mack, Thomas Heimbucher, and Coleen T. Murphy, “The nematode Caenorhabditis elegans as a model for aging research,” Drug Discovery Today: Disease Models 27, 2018, pp. 3-13, doi: 10.1016/j.ddmod.2018.11.001.
[58] Jeong D.E. et al., “Inhibition of the oligosaccharyl transferase in Caenorhabditis elegans that compromises ER proteostasis suppresses p38-dependent protection against pathogenic bacteria,” PLoS Genet 16(3), 2020, doi: 10.1371/journal.pgen.1008617.