阿茲海默症是一種失能症，能對人記憶功能，思考行為及日常活動造成嚴重問題。阿茲海默症是種進展性疾病，會隨著時間而變得更加嚴重，並且使老年人的生活水準顯著下降，造成嚴重的社會問題。阿茲海默症目前並無治療法，只能用藥物來緩解症狀。由於此病早期的症狀和正常老化有些相似，許多人直到病情進展到輕度認知受損期才發現。
雖然臨床診斷依舊是阿茲海默症診斷的黃金標準，我們還是盡量尋找阿茲海默症的生物特徵以減輕專家們的負擔。核磁共振影像是不錯的出發點，因為其在能察覺灰質變化的影像方法中是最多機構使用的。
本文的目的在設計一個可以自動偵測敏感區域、特徵擷取及最終分類的分類流程。分類流程中使用許多深度卷積網路，其在視覺學習領域算是最普遍的方法，此外也會探討神經科學常用的影像特徵對此分類器的幫助。經由我們的處理流程，我們可以正確分析出海馬廻所在區域最能代表敏感區域，並且海馬廻、小腦、枕葉及額葉都成功的被截取為判別阿茲海默症的特徵。最後的判斷準確度在無先驗知識下約為92.6％。

Alzheimer’s disease (AD) is a type of dementia that causes trouble to memory function, thinking process and daily activities. AD is a progressive disease, getting worse and worse over time. It seriously hurts the quality of life of elderly adults, becoming an important social problem. There is no current cure for AD, only drug treatment for symptom control. Since symptoms of AD are like normal aging at early stage, most patients do not aware of the symptoms until mild cognitive impairment (MCI) stage.
Although the clinical diagnosis criteria of AD are still the golden standard in AD detection, we also seek some representative biomarkers to alleviate expert’s burdens. MRI is a good starting point, since it is the most popular imaging method that could detect gray matter change.
The goal of this article is to design a classification procedure that do ROI selection, feature extraction and final classification automatically. During learning procedure, we intensively use convolutional neural networks, which is the most popular training method in visual learning fields. Besides, we also investigate if popular image processing procedure in neuroscience could benefit deep learning architectures. Follow our processing procedure, models could accurately learn that slices contain hippocampus are the most likely region of interest. In feature extraction part, networks could correctly extract features like hippocampus, cerebellum, frontal lobe and occipital lobe as Alzheimer’s disease features. The final classification accuracy is about 92.6％ without any prior knowledge.

Table of Contents
Chinese Abstract ------------------------------------- i
English Abstract ------------------------------------- ii
Table of Contents ----------------------------------- iv
List of Figures --------------------------------------- vi
List of Tables ---------------------------------------- vii
Chapter 1 Introduction -------------------------- 1
1.1 Alzheimer’s disease introduction-------------- 1
1.2 Biomarkers of Alzheimer’s disease------------- 3
1.3 Machine Learning Flow ------------------------ 4
1.4 Neural Network ------------------------------- 6
1.5 Convolution Neural Network ------------------- 8
1.6 Frame of this thesis ------------------------ 11
Chapter 2 Structural pipeline ------------------ 13
2.1 Dataset description ------------------------- 13
2.2 Subjects ------------------------------------ 13
2.3 Data acquisition protocol-------------------- 15
2.4 Structural T1 image pipeline------------------17
Chapter 3 Classification pipeline -------------- 21
3.1 Classification overview --------------------- 21
3.2 Slice selection – Caffenet ------------------ 22
3.3 Slice feature learning – ResNet ------------- 29
3.4 Output selection – Linear SVM --------------- 31
3.4 Computational environment setting ----------- 33
Chapter 5 Discussion --------------------------- 34
Chapter 6 Conclusion --------------------------- 36
References -------------------------------------- 37


List of Figures
Fig 1. Traditional machine learning flow ----------- 4
Fig 2. Deep learning flow -------------------------- 5
Fig 3. Artificial neuron state transition diagram --- 7
Fig 4. Feed-forward network diagram ---------------- 7
Fig 5. Example LeNet 5 processing procedure ---------- 9
Fig 6. Structural pipeline overview ----------------- 20
Fig 7. Overall classification procedure ----------- 22
Fig 8. Caffenet architecture ---------------------- 23
Fig 9. Error comparison in x axis ----------------- 26
Fig 10. Error comparison in y axis ------------------ 27
Fig 11. Error comparison in z axis ---------------- 28
Fig 12. First 5 layers in Resnet – 34 -------------- 30
Fig 13. Activation map and learned features --------- 31

List of Tables
Table 1. Siemens 3.0 T acquiring parameters -------- 15
Table 2. Philips 3.0 T acquiring parameters -------- 16
Table 3. Accuracy of slice classification ---------- 32

Reference
Apostolova, L. G., & Thompson, P. M. (2008). Mapping progressive brain structural changes in early Alzheimer's disease and mild cognitive impairment. Neuropsychologia, 46(6), 1597-1612.

Ashburner, J., & Friston, K. J. (2000). Voxel-based morphometry—the methods. Neuroimage, 11(6), 805-821.

Delacourte, A., & Defossez, A. (1986). Alzheimer's disease: Tau proteins, the promoting factors of microtubule assembly, are major components of paired helical filaments. Journal of the neurological sciences, 76(2), 173-186.

Deng, L., Li, J., Huang, J. T., Yao, K., Yu, D., Seide, F., ... & Gong, Y. (2013, May). Recent advances in deep learning for speech research at Microsoft. In 2013 IEEE International Conference on Acoustics, Speech and Signal Processing (pp. 8604-8608). IEEE.

Douaud, G., Jbabdi, S., Behrens, T. E., Menke, R. A., Gass, A., Monsch, A. U., ... & Smith, S. (2011). DTI measures in crossing-fibre areas: increased diffusion anisotropy reveals early white matter alteration in MCI and mild Alzheimer's disease. Neuroimage, 55(3), 880-890.

He, K., Zhang, X., Ren, S., & Sun, J. (2015). Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. In Proceedings of the IEEE International Conference on Computer Vision (pp. 1026-1034).

Hyman, B. T., Van Hoesen, G. W., Damasio, A. R., & Barnes, C. L. (1984). Alzheimer's disease: cell-specific pathology isolates the hippocampal formation. Science, 225(4667), 1168-1170.

Ilonen, J., et al. (2003). "Differential evolution training algorithm for feed-forward neural networks." Neural Processing Letters 17(1): 93-105.

Ioffe, S., & Szegedy, C. (2015). Batch normalization: Accelerating deep network training by reducing internal covariate shift. arXiv preprint arXiv:1502.03167.

Jenkinson, M., Pechaud, M., & Smith, S. (2005, June). BET2: MR-based estimation of brain, skull and scalp surfaces. In Eleventh annual meeting of the organization for human brain mapping (Vol. 17, p. 167).

Jia, Y., Shelhamer, E., Donahue, J., Karayev, S., Long, J., Girshick, R., ... & Darrell, T. (2014, November). Caffe: Convolutional architecture for fast feature embedding. In Proceedings of the 22nd ACM international conference on Multimedia (pp. 675-678). ACM.

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).

Le, Q. V., Zou, W. Y., Yeung, S. Y., & Ng, A. Y. (2011, June). Learning hierarchical invariant spatio-temporal features for action recognition with independent subspace analysis. In Computer Vision and Pattern Recognition (CVPR), 2011 IEEE Conference on (pp. 3361-3368). IEEE.

LeCun, Y., et al. (1995). Comparison of learning algorithms for handwritten digit recognition. International conference on artificial neural networks.

McKhann, G. M., Knopman, D. S., Chertkow, H., Hyman, B. T., Jack, C. R., Kawas, C. H., ... & Mohs, R. C. (2011). The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimer's & dementia, 7(3), 263-269.

Mei, P. A., de Carvalho Carneiro, C., Fraser, S. J., Min, L. L., & Reis, F. (2015). Analysis of neoplastic lesions in magnetic resonance imaging using self-organizing maps. Journal of the neurological sciences, 359(1), 78-83.

Merriam, A. E., Aronson, M. K., Gaston, P., Wey, S. L., & Katz, I. (1988). The psychiatric symptoms of Alzheimer's disease. Journal of the American Geriatrics Society, 36(1), 7-22.

Nesterov, Y., & Nemirovskii, A. (1994). Interior-point polynomial algorithms in convex programming (Vol. 13). Siam.

Nestor, S. M., Rupsingh, R., Borrie, M., Smith, M., Accomazzi, V., Wells, J. L., ... & Alzheimer's Disease Neuroimaging Initiative. (2008). Ventricular enlargement as a possible measure of Alzheimer's disease progression validated using the Alzheimer's disease neuroimaging initiative database. Brain, 131(9), 2443-2454.

Payan, A., & Montana, G. (2015). Predicting Alzheimer's disease: a neuroimaging study with 3D convolutional neural networks. arXiv preprint arXiv:1502.02506.

Scheuner, D., Eckman, C., Jensen, M., Song, X., Citron, M., Suzuki, N., ... & Larson, E. (1996). Secreted amyloid beta-protein similar to that in the senile plaques of Alzheimer's disease is increased in vivo by the presenilin 1 and 2 and APP mutations linked to familial Alzheimer's disease. Nature medicine, 2(8), 864-870.

Sharif Razavian, A., Azizpour, H., Sullivan, J., & Carlsson, S. (2014). CNN features off-the-shelf: an astounding baseline for recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (pp. 806-813).

Shattuck, D. W., Mirza, M., Adisetiyo, V., Hojatkashani, C., Salamon, G., Narr, K. L., ... & Toga, A. W. (2008). Construction of a 3D probabilistic atlas of human cortical structures. Neuroimage, 39(3), 1064-1080.

Srivastava, N., Hinton, G. E., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R. (2014). Dropout: a simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 15(1), 1929-1958.

Wang, L., Beg, F., Ratnanather, T., Ceritoglu, C., Younes, L., Morris, J. C., ... & Miller, M. I. (2007). Large deformation diffeomorphism and momentum based hippocampal shape discrimination in dementia of the Alzheimer type. IEEE transactions on medical imaging, 26(4), 462-470.

Yang, S. T., Lee, J. D., Chang, T. C., Huang, C. H., Wang, J. J., Hsu, W. C., ... & Li, K. Y. (2013). Discrimination between Alzheimer's disease and mild cognitive impairment using SOM and PSO-SVM. Computational and mathematical methods in medicine, 2013.

Yoshita, M., Fletcher, E., Harvey, D., Ortega, M., Martinez, O., Mungas, D. M., ... & DeCarli, C. S. (2006). Extent and distribution of white matter hyperintensities in normal aging, MCI, and AD. Neurology, 67(12), 2192-2198.

Zhang, Y., Brady, M., & Smith, S. (2001). Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE transactions on medical imaging, 20(1), 45-57.