阿茲海默氏症為一多重病徵之疾病，其症狀包含記憶力衰退，並伴隨著其他認知功能之退化；除此之外，病患甚至會出現妄想、幻覺、性格大變等行為問題。
在本研究中，期望可以藉由將遷移學習運用至卷積神經網路，協助醫生在使用核磁共振影像診斷阿茲海默症時，能更有效且更精準地做辨別。
在這項研究中，將使用著名圖像分類卷積神經網路VGGNet，且已在大數據ImageNet上完成預訓練。僅將原始VGGNet的全連接層剃除，結合本實驗所設計之全連接層。
最終結果顯示，在分類是否患有阿茲海默氏症之準確率可達0.76。然而，在對阿茲海默氏症等級的三分類上，準確率僅為0.2，由此表示本研究中的微調VGG16，無法區別出不同等級的阿茲海默氏症。

Alzheimer's disease is a multiple symptoms disease. It’s the most common cause of dementia ─ a continuous decline in thinking, behavioral and social skills.A person with Alzheimer's disease will develop severe memory impairment and the deterioration of other cognitive functions. There may even be delusions, hallucinations, or disruptive behaviors, and some will change their personality.
In this study, we expect that transfer learning can be applied to convolutional neural networks to assist doctors diagnosed Alzheimer's disease more effectively and accurately by using MRI.
In this research, we used the well-known image classification convolutional neural network VGGNet, and completed pre-training on the big data ImageNet. Then using fully connected layer only cut from original VGGNet and combine the fully connected layer of my own design.
Under optimal conditions, the results show that the proposed approach achieved the accuracy of 0.76 for Alzheimer's disease classification. In addition ,the accuracy of three levels classification of Alzheimer's disease is only 0.2. In summary, different levels of Alzheimer's disease cannot be distinguished by fine-tuning of VGG16 in this study.

1. 前言 17
2. 原理 23
2.1. 類神經網路 23
2.2. 類神經網路的運作原理 24
2.2.1. 前向傳播(Forward Propagation) 24
2.2.2. 反向傳播(Back Propagation) 25
2.2.3. 梯度消失(Vanishing Gradient) 30
2.3. 類神經網路的優化 31
2.3.1. 限制玻爾茲曼機(Restricted Boltzmann Machines, RBM) 32
2.3.2. 深度信念網路(Deep Belief Networks, DBN) 36
2.4. 激活函數(Activation Function) 39
2.4.1. 線性分類 39
2.4.2. 非線性分類 41
2.5. 機器學習(Machine Learning, ML) 43
2.5.1. 監督式學習(Supervised Learning) 44
2.5.2. 非監督式學習(Unsupervised Learning) 44
2.5.3. 強化式學習(Reinforcement Learning) 45
2.6. 深度學習(Deep Learning, DL) 46
2.6.1. 卷積神經網路(Convolutional Neural Network, CNN)的起源 46
2.6.2. 神經感知機(Neocognitron) 48
2.6.3. 卷積神經網路模型 49
2.7. 遷移學習(Transfer Learning, TL) 51
2.7.1. 樣本遷移(Instance Based TL) 51
2.7.2. 特徵遷移(Feature Based TL) 52
2.7.3. 模型遷移(Parameter Based TL) 52
2.7.4. 關係遷移(Relation Based TL) 52
3. 深度學習的模型原理 53
3.1. 卷積層(Convolution Layer) 59
3.2. 池化層(Pooling Layer) 61
3.3. ReLU激活函數 63
3.4. 全連接層(Fully Connection, FC)中的平坦層 65
3.5. 全連接層(Fully Connection, FC)中的隱藏層 66
3.6. Dropout 67
3.7. 全連接層中(Fully Connection, FC)的輸出層 68
4. 數據處理 70
4.1. 資料庫介紹 70
4.2. 影像挑選 72
4.3. 影像預處理(Image Processing) 77
4.3.1. 內插法(Interpolation) 77
4.3.2. 背景尺寸(Canvas Size) 79
4.3.3. 切片重組(Reslice) 79
4.3.4. 正規化(Nomalization) 80
4.3.5. 資料擴增(Data Augmentation) 81
5. 實驗設計 86
5.1. 樣本準備 86
5.1.1. 樣本設計 87
5.1.2. 切分樣本 88
5.2. 模型輸入層設計 95
5.3. 建立模型 95
5.3.1. 微調模型 95
5.3.2. 演算法 98
5.4. 模型訓練 99
5.4.1. 超參數定義 99
5.4.2. 超參數設定 99
5.5. 評估模型 100
5.5.1. 混淆矩陣 100
5.5.2. ROC曲線 102
6. 實驗結果 104
6.1. 正常對照組和已確診阿茲海默氏症 104
6.2. 正常對照組、輕度阿茲海默氏症與中重度阿茲海默氏症 109
6.3. 正常對照組與輕度阿茲海默氏症 114
6.4. 輕度阿茲海默氏症與中重度阿茲海默氏症 121
6.5. 正常對照組與中重度阿茲海默氏症 128
7. 討論 135
8. 結論與未來展望 140

[1] R.E. Tanzi, A Brief History of Alzheimer's Disease Gene Discovery, Journal of Alzheimer's Disease, 33 (2013) S5-S13.
[2] D.J. Selkoe, Alzheimer's Disease: Genes, Proteins, and Therapy, Physiological Reviews, 81 (2001) 741-766.
[3] R.F. Allegri, D. Sarasola, C.M. Serrano, F.E. Taragano, R.L. Arizaga, J. Butman, L. Loñ, Neuropsychiatric symptoms as a predictor of caregiver burden in Alzheimer's disease, Neuropsychiatr Dis Treat, 2 (2006) 105-110.
[4] N.C. Fox, E.K. Warrington, P.A. Freeborough, P. Hartikainen, A.M. Kennedy, J.M. Stevens, M.N. Rossor, Presymptomatic hippocampal atrophy in Alzheimer's disease: A longitudinal MRI study, Brain, 119 (1996) 2001-2007.
[5] S.A.R.B. Rombouts, F. Barkhof, M.P. Witter, P. Scheltens, Unbiased whole-brain analysis of gray matter loss in Alzheimer's disease, Neuroscience Letters, 285 (2000) 231-233.
[6] L.T. Westlye, K.B. Walhovd, A.M. Dale, T. Espeseth, I. Reinvang, N. Raz, I. Agartz, D.N. Greve, B. Fischl, A.M. Fjell, Increased sensitivity to effects of normal aging and Alzheimer's disease on cortical thickness by adjustment for local variability in gray/white contrast: A multi-sample MRI study, NeuroImage, 47 (2009) 1545-1557.
[7] K. Juottonen, M.P. Laakso, K. Partanen, H. Soininen, Comparative MR Analysis of the Entorhinal Cortex and Hippocampus in Diagnosing Alzheimer Disease, American Journal of Neuroradiology, 20 (1999) 139.
[8] A.S. Fleisher, S. Sun, C. Taylor, C.P. Ward, A.C. Gamst, R.C. Petersen, C.R. Jack, P.S. Aisen, L.J. Thal, Volumetric MRI vs clinical predictors of Alzheimer disease in mild cognitive impairment, Neurology, 70 (2008) 191.
[9] A.S. Lundervold, A. Lundervold, An overview of deep learning in medical imaging focusing on MRI, Zeitschrift für Medizinische Physik, 29 (2019) 102-127.
[10] M. Deepika Nair, M.S. Sinta, M. Vidya, A Study on Various Deep Learning Algorithms to Diagnose Alzheimer’s Disease, in: D. Pandian, X. Fernando, Z. Baig, F. Shi (Eds.) Proceedings of the International Conference on ISMAC in Computational Vision and Bio-Engineering 2018 (ISMAC-CVB), (Springer International Publishing, Cham, 2019), pp. 1705-1710.
[11] S.G. Mueller, M.W. Weiner, L.J. Thal, R.C. Petersen, C.R. Jack, W. Jagust, J.Q. Trojanowski, A.W. Toga, L. Beckett, Ways toward an early diagnosis in Alzheimer’s disease: The Alzheimer’s Disease Neuroimaging Initiative (ADNI), Alzheimer's & Dementia, 1 (2005) 55-66.
[12] G.a. Chetelat, J.-C. Baron, Early diagnosis of alzheimer’s disease: contribution of structural neuroimaging, NeuroImage, 18 (2003) 525-541.
[13] R. Cuingnet, E. Gerardin, J. Tessieras, G. Auzias, S. Lehéricy, M.-O. Habert, M. Chupin, H. Benali, O. Colliot, Automatic classification of patients with Alzheimer's disease from structural MRI: A comparison of ten methods using the ADNI database, NeuroImage, 56 (2011) 766-781.
[14] E. Gerardin, G. Chételat, M. Chupin, R. Cuingnet, B. Desgranges, H.-S. Kim, M. Niethammer, B. Dubois, S. Lehéricy, L. Garnero, F. Eustache, O. Colliot, Multidimensional classification of hippocampal shape features discriminates Alzheimer's disease and mild cognitive impairment from normal aging, NeuroImage, 47 (2009) 1476-1486.
[15] H.-I. Suk, S.-W. Lee, D. Shen, Hierarchical feature representation and multimodal fusion with deep learning for AD/MCI diagnosis, NeuroImage, 101 (2014) 569-582.
[16] O. Querbes, F. Aubry, J. Pariente, J.-A. Lotterie, J.-F. Démonet, V. Duret, M. Puel, I. Berry, J.-C. Fort, P. Celsis, I. The Alzheimer's Disease Neuroimaging, Early diagnosis of Alzheimer's disease using cortical thickness: impact of cognitive reserve, Brain, 132 (2009) 2036-2047.
[17] L. Sørensen, C. Igel, A. Pai, I. Balas, C. Anker, M. Lillholm, M. Nielsen, I. Alzheimer’s Disease Neuroimaging, B. the Australian Imaging, a. Lifestyle flagship study of, Differential diagnosis of mild cognitive impairment and Alzheimer's disease using structural MRI cortical thickness, hippocampal shape, hippocampal texture, and volumetry, Neuroimage Clin, 13 (2016) 470-482.
[18] S. Basaia, F. Agosta, L. Wagner, E. Canu, G. Magnani, R. Santangelo, M. Filippi, Automated classification of Alzheimer's disease and mild cognitive impairment using a single MRI and deep neural networks, NeuroImage: Clinical, 21 (2019) 101645.
[19] J. Liu, M. Li, W. Lan, F. Wu, Y. Pan, J. Wang, Classification of Alzheimer's Disease Using Whole Brain Hierarchical Network, IEEE/ACM Transactions on Computational Biology and Bioinformatics, 15 (2018) 624-632.
[20] S.J. Pan, Q. Yang, A Survey on Transfer Learning, IEEE Transactions on Knowledge and Data Engineering, 22 (2010) 1345-1359.
[21] M. Hon, N.M. Khan, Towards Alzheimer's disease classification through transfer learning, in: 2017 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), (2017), pp. 1166-1169.
[22] A. Ebrahimi-Ghahnavieh, S. Luo, R. Chiong, Transfer Learning for Alzheimer's Disease Detection on MRI Images, in: 2019 IEEE International Conference on Industry 4.0, Artificial Intelligence, and Communications Technology (IAICT), (2019), pp. 133-138.
[23] B. Khagi, C.G. Lee, G.-R. Kwon, Alzheimer’s disease Classification from Brain MRI based on transfer learning from CNN, in: 2018 11th Biomedical Engineering International Conference (BMEiCON), (IEEE, 2018), pp. 1-4.
[24] K. Zhou, W. He, Y. Xu, G. Xiong, J. Cai, Feature selection and transfer learning for Alzheimer’s disease clinical diagnosis, Applied Sciences, 8 (2018) 1372.
[25] G.E. Hinton, R.R. Salakhutdinov, Reducing the Dimensionality of Data with Neural Networks, Science, 313 (2006) 504.
[26] G.E. Hinton, S. Osindero, Y.-W. Teh, A Fast Learning Algorithm for Deep Belief Nets, Neural Computation, 18 (2006) 1527-1554.
[27] W.S. McCulloch, W. Pitts, A logical calculus of the ideas immanent in nervous activity, The bulletin of mathematical biophysics, 5 (1943) 115-133.
[28] D.H. Hubel, T.N. Wiesel, Receptive fields of single neurones in the cat's striate cortex, J Physiol, 148 (1959) 574-591.
[29] Y. Lecun, L. Bottou, Y. Bengio, P. Haffner, Gradient-based learning applied to document recognition, Proceedings of the IEEE, 86 (1998) 2278-2324.
[30] A. Krizhevsky, I. Sutskever, G.E. Hinton, Imagenet classification with deep convolutional neural networks, in: Advances in neural information processing systems, (2012), pp. 1097-1105.
[31] C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V. Vanhoucke, A. Rabinovich, Going deeper with convolutions, in: Proceedings of the IEEE conference on computer vision and pattern recognition, (2015), pp. 1-9.
[32] K. Simonyan, A. Zisserman, Very deep convolutional networks for large-scale image recognition, arXiv preprint arXiv:1409.1556, (2014).
[33] K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, in: Proceedings of the IEEE conference on computer vision and pattern recognition, (2016), pp. 770-778.
[34] G.E. Hinton, N. Srivastava, A. Krizhevsky, I. Sutskever, R.R. Salakhutdinov, Improving neural networks by preventing co-adaptation of feature detectors, arXiv preprint arXiv:1207.0580, (2012).
[35] H. Su, S. Maji, E. Kalogerakis, E. Learned-Miller, Multi-view convolutional neural networks for 3d shape recognition, in: Proceedings of the IEEE international conference on computer vision, (2015), pp. 945-953.
[36] N. Amoroso, D. Diacono, A. Fanizzi, M. La Rocca, A. Monaco, A. Lombardi, C. Guaragnella, R. Bellotti, S. Tangaro, Deep learning reveals Alzheimer's disease onset in MCI subjects: Results from an international challenge, Journal of Neuroscience Methods, 302 (2018) 3-9.
[37] C. Studholme, D.L.G. Hill, D.J. Hawkes, An overlap invariant entropy measure of 3D medical image alignment, Pattern Recognition, 32 (1999) 71-86.
[38] M. Maqsood, F. Nazir, U. Khan, F. Aadil, H. Jamal, I. Mehmood, O.-y. Song, Transfer Learning Assisted Classification and Detection of Alzheimer’s Disease Stages Using 3D MRI Scans, Sensors, 19 (2019) 2645.