核磁共振技術(Magnetic Resonance Imaging, MRI)為現今臨床上重要的檢測技術，其核磁共振技術最大優點是對人體不具侵襲性，且可以多方向掃描，並提供三度空間、高對比度的影像。可有利於醫師對疾病的診斷更加準確，以提高治療的正面效果。當一病人被安排做MRI造影後，對人體某一器官切面部位，會產生一系列的多頻譜影像(multi-spectral image)。如果把一系列的切面影像疊在一起組合而成後，便形成人體一個立體的三維結構。醫師就藉由這個立體結構得到一些醫學診斷的資訊，如器官的形狀、位置與體積大小。雖然經由多頻譜影像可獲得更多的資訊，但也造成病理判讀上的困擾。因此，我們將這些多頻譜影像經過精準的轉換法處理後，形成單一強化組織影像讓醫生更容易的對病理做診斷。
此篇論文提出了一個新特徵自我選取的方法，Target generation process(TGP)。並將TGP合併於Linear Discriminant Analysis (LDA) 與 Support Vector Machine (SVM)兩方法。我們稱此兩方法為Unsupervised Linear Discriminant Analysis (ULDA) 與 Unsupervised Support Vector Machine (USVM)。利用ULDA(或USVM)來強化出腦中的CSF(cerebrospinal fluid),白質(White Matter)以及灰質(Gray Matter)三大組織，使醫生做診斷時更加有效率。因此我們的工作即在研究如何從多頻譜MRI影像中，將腦部的主要組織(如CSF、WM、GM)給強化出來，且亦有一套方法來評比這些方法的可行性與強健性。

Magnetic Resonance Imaging (MRI) is a useful medical instrument in medical science. It provides unparallel capability of revealing soft tissue characterization as well as 3-D visualization and proposes the diagnosis without needing to intrude into the human body. MRI produces a sequence of multiple spectral images of tissues with a variety of contrasts, but the multi-spectral images cannot be conveniently used to be a pathology diagnosis correctly. In general, we need to transform the multi-spectral images to an enhanced image which is easier to be used for doctor’s clinical diagnosis. One of the potential applications of MRI in clinical practice is the brain parenchyma classification.
In this thesis, we first present a feature selection method called Target Generation Process (TGP) and the TGP generates a set of potential targets from an unknown background. Let the targets be the training data for the classifiers of Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM), respectively, such that the classification and segmentation for the MR images are achieved. The algorithm combining TGP and LDA (or SVM) is called Unsupervised Linear Discriminant Analysis (ULDA) (or Unsupervised Support Vector Machine (USVM)). Finally, the effectiveness of ULDA and USVM in target classification is evaluated by several MRI images experiments. In order to further evaluate its performance, they are compared with the method of Fuzzy c-mean. Several experiment results show that the ULDA and USVM have the better effective segmentation for multi-spectral MR images.

[1] C. M. Wang, et al., “An Unsupervised Orthogonal Subspace Projection Approach to MR Image Classification," Optical Engineering, vol. 41, no. 7, pp. 1546-1557, July. 2002
[2] C. M. Wang, et al., "Orthogonal Subspace Projection-Base Approaches to Classification of MR Images Sequences," Computerized Medical Imaging and Graphics, vol. 25, no. 6, pp. 465-476, December. 2001.
[3] C. M. Wang, et al., “Detection of Spectral Signatures in Multispectral MR Images for Classification,” IEEE Transactions on Medical Imaging, vol. 22, pp.50 – 61, January 2003
[4] C. M. Wang, et al., “An Unsupervised Kalman Filter-based Linear Mixing Approach to MRI Classification,” Proceedings of the 2004 IEEE Asia-Pacific Conference on Circuits and Systems, vol.2, pp.1105 – 1108, 6-9 December 2004.
[5] W. E.Reddick, et al., "Automated Segmentation and Classification of Multispectral Magnetic Rresonance Images of Brain using Artificial Neural Networks," IEEE Transactions on Medical Imaging, vol. 16, no. 6, pp. 911-918, December 1997.
[6] J. Alirezaie, M. E. Jernigan, and C. Nahmias "Automatic Segmentation of Cerebral MR Images using Artificial Neural Networks," IEEE Transactions on Nuclear Science, vol. 45, no. 4, pp. 2174-2182, August 1998
[7] J. S. Lin, R. M. Chen, Y. M. Huang, "Medical Image Segmentation using Mean Field Annealing Network," IEEE International Conference on Image Processing, vol. 2, pp. 855-858, October 1997
[8] J. S. Lin, K. S. Cheng, C. W. Mao, "Multispectral Magnetic Resonance Images Segmentation using Fuzzy Hopfield Neural Network," International Journal of Biomedical Computing, pp. 205-214, August 1996.
[9] M. N. Ahmed, S et al., “A Modified Fuzzy C-means Algorithm for Bias Field Estimation and Segmentation of MRI Data,” IEEE Transactions on Medical Imaging, vol. 21, pp.193 – 199, March 2002.
[10] K. H. Chuang, et al., “Model-free Functional MRI Analysis using Kohonen Clustering Neural Network and Fuzzy C-means,” IEEE Transactions on Medical Imaging, vol. 18, pp.1117 – 1128, December 1999
[11] Y. Zhao and M. Li, “A Modified Fuzzy C-means Algorithm for Segmentation of MRI,” Proceedings of the Fifth International Conference on Computational Intelligence and Multimedia Applications, pp.391 – 395, 27-30 September 2003
[12] M. N. Ahmed, et al., “Bias Field Estimation and Adaptive Segmentation of MRI Data using a Modified Fuzzy C-means Algorithm,” IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Vol. 1, 23-25 June 1999
[13] N. A. Mohamed, M. N. Ahmed, and A. Farag, “Modified Fuzzy C-mean in Medical Image Segmentation,” 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing, vol.6, pp.3429 – 3432, 15-19 March 1999
[14] M. C. Clark, et al., “Automatic Tumor Segmentation using Knowledge-based Techniques,” IEEE Transactions on Medical Imaging, vol.17 pp.187-201, 1998
[15] W. S. Yambor, “Analysis of PCA-based and Fisher Discriminant-Based Image Recognition Algorithms,” M.S. Thesis, Technical Report CS-00-103, Computer Science Department, Colorado State University, July 2000
[16] K. Etemad and R. Chellappa, “Discriminant Analysis for Recognition of Human Face Images,” Journal of the Optical Society of America A, Vol. 14, pp. 1724-1733, August 1997
[17] J. W. Hung and L. S Lee, “Optimization of Temporal Filters for Constructing Robust Features in Speech Recognition,” IEEE Transactions on Audio, Speech and Language Processing, Vol. 14, pp.808 - 832, May 2006
[18] P. Howland, M. Jeon, and H. Park, “Structure Preserving Dimension Reduction for Clustered Text Data Based on the Generalized Singular Value Decomposition,” SIAM Journal on Matrix Analysis and Applications, Vol. 25, pp.165-179 , 2003
[19] C. Cortes, V. Vapnik, "Support Vector Networks," Machine Learning, vol.20, pp. 273-297, 1995.
[20] S. Mika. ‘Kernel Fisher Discriminants,” PhD Thesis, Technische Universit