在腦部疾病的診斷上，醫學影像已在臨床上被廣泛地使用，核磁共振造影(MRI)的問世更使影像醫學又邁進一個新的紀元，核磁共振之造影方式是非侵入性的，臨床的診斷中，MRI使用三個參數：T1分量、T2分量及雙重回應的質子密度(PD)，產生不同對比度的free water、proteinaceous fluid、軟組織及其他組織等影像，由於提供多個頻譜之豐富的資訊，MRI對於臨床醫師在診斷上，比其他形式的影像來得有用許多，在臨床的診斷上，腦部正常與病態組織的分割及分類，是最重要也是首要的步驟，經由分類及分割出之腦組織的體積、形狀及分佈區域等資訊，可以進一步地找出不正常的區域及病灶，諸如皮質異位(heterotopia)、平腦(lissencephaly)、腦萎縮(brain atrophy)以及大腦梗塞(cerebral infarction)等。
MRI影像進行組織分類時，由於影像是多頻譜影像，因此如果只以某特定頻譜的單張影像，進行處理，會因單張影像資訊不足，無法達到組織分類的目的，所以我們必須結合所有頻譜影像的資訊來做組織的分類，因此使用多頻譜影像處理技術，擷取各頻譜資訊進行分類，在臨床上有著相當重要的價值。在從事多頻譜影像處理技術於核磁共振腦部造影分類之研究過程中，我們建構出其分析模型，進行一序列的模擬與實驗，進而推導出我們所要的結果。在這研究過程中，正交次空間投影法(Orthogonal Subspace Projection, OSP)、非監督式正交次空間投影法(Unsupervised OSP) 及抑制能量最小化法(Constrained Energy Minimization, CEM)均發表論文去展現出我們的結果。
正交次空間投影法(Orthogonal Subspace Projection)，這個方法的基本概念是將影像中物件信號分成期望與非期望信號，藉由投射信號的向量到頻譜特徵正交次空間(subspace)，達到抑制非期望(undesired) 信號，接著利用比對濾波器偵測出期望的信號，然而OSP需要求完整的物質背景知識，為了使OSP更具實用與效率，我們進一步擴展OSP成為Unsupervised OSP，使得 OSP 在實際的應用不需要完全的影像背景資料，這個理論讓OSP在MRI的應用上更具有吸引力。另外一個方法是抑制能量最小化法(CEM)，其具有UOSP 之相同優點，CEM所使用的基礎，就是矩陣運算中的最小化變異數無雜訊響應(Minimum Variance Distortionless Response, MVDR)，CEM將多頻譜MR影像視為一個矩陣處理的問題，在這問題中的每一個觀察值都代表一個頻譜頻帶(Spectral Band)，再使用有限脈衝響應濾波器(Finite Impulse Response ,FIR)將輸出功率做最小化處理，使期望的特徵被強制到一個特定的增益。為了印證前面所提出的三個方法，我們進行兩類的實驗進行效能分析，其一是使用電腦模擬影像（Phantom），進行量化之研究，另一類實驗則使用取自台中榮總之實際核磁共振造影，經由實驗結果，顯示出大腦組織被正確地分割出灰質(Gray Matter)、白質(White Matter)以及腦脊髓液(Cerebral Spinal Fluid ,CSF)，而在量化方面之研究比較，也很明確顯示出其優於C-Means這個方法所呈現的結果。

Magnetic Resonance Imaging (MRI) has become a widely used modality because it produces multispectral image sequences that provide information of free water, proteinaceous fluid, soft tissue and other tissues with a variety of contrast. The abundance fractions of tissue signatures provided by multispectral images can be very useful for medical diagnosis compared to other modalities. In particular, brain parenchyma classification and segmentation of normal and pathological tissue is the first step of addressing a wide range of clinical problems. Using information of volumes, shapes and region distributions of brain tissues, one can find abnormalities that are commonly related to heterotopia, lissencephaly, brain atrophy, and cerebral infarction. In this dissertation we investigate applications of three multispectral image processing techniques to MRI of brain classification, which are Orthogonal Subspace Projection (OSP), Unsupervised OSP (UOSP) and Constrained Energy Minimization (CEM).
The OSP has been shown great success in remote sensing image processing. It divides object signatures into desired and undesired object signatures with the latter annihilated by orthogonal projection prior to detection of the former by a matched filter. Since the OSP requires complete knowledge of object signatures, which is generally difficult to obtain in practice, it is further extended to an unsupervised OSP in the sense that only the knowledge of the desired object signatures is required and the undesired object signatures can be generated directly from image data in an unsupervised manner.
Constrained Energy Minimization (CEM) is an alternative to the UOSP, but adopts a completely different approach. It is derived from the Minimum Variance Distortionless Response (MVDR) arising in array processing. It does not need the knowledge of undesired object signatures as does the UOSP. It uses a Finite Impulse Response (FIR) filter to minimize the filter output power while extracting the desired object signatures using linear constraints. The CEM has an advantage over the UOSP in that it does not require the knowledge of image background. Such advantage is very useful for MR image classification where many unknown signal sources may be present in image data resulting from varying structures of human issues that cannot be characterized and identified by visual inspection a priori.
In order to evaluate the three proposed methods, two types of experiments are conducted for performance analysis. The first set of data consists of phantom brain images that can be used for a quantitative study. The second set of data is made up of real MR brain images that were acquired from ten patients with normal brain function in Taichung Veterans General Hospital. The experimental results have demonstrated that the three proposed remote sensing image processing techniques perform significantly better than the commonly used C-Means method.

Bibliography
[1] B. Johnston, M. S. Atkins, B. Mackiewich, and M. Anderson, “Segmentation of multiple sclerosis lesions in intensity corrected multispectral MRI,” IEEE Transactions on Medical Imaging, vol. 15, no. 2, pp. 154-169, April 1996.
[2] Atam P. Dhawan, Aleksandar Zavaljevski, Alok Sarwal, Walliam S. Ball, “A system for MR brain image segmentation,” 18th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Amsterdam, pp. 732-733, 1996.
[3] Laurent Verard, Jalal Fadidi, Su Ruan, and Daniel Bloyet, “3D MRI segmentation of brain structures,” 18th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Amsterdam, pp. 1081-1082, 1996.
[4] H. Grahn, N.M. Szeverenyi, N.W. Roggenbuck, F. Delaglio and P. Geladi, “Data analysis of multivariate magnetic resonance images I. a principal component analysis approach,” Chemometrics and Intelligent Lab. Systems, vol. 5, pp. 11-322, 1989.
[5] J.P. Windham, M.A. Abd-Allah, D.A. Reimann, J.W. Froehich and A.M. Haggar, “Eigneimage filtering in MR imaging,“ J. Computer Assisted Tomography, vol. 12, no. 1, pp. 1-9, Jan/Feb. 1988.
[6] H. Soltanian-Zadeh and J.P. Windham, “Novel and general approach to linear filter designed for contrast-to-noise ratio enhancement of magnetic resonance images with multiple interfering features in the scene,” J. Electronic Imaging, vol. 1, no. 2, pp. 171-182, April 1992.
[7] A.M. Haggar, J.P. Windham D.A. Reimann, D.C. Hearshen and J.W. Froehich, “Eigneimage filtering in MR imagine: an application in the abnormal chest wall,” Magnetic Resonance in Medicine, vol. 11, pp. 85-97 1989.
[8] H. Soltanian-Zadeh, J.P. Windham, H. Soltanian-Zadeh, J.P. Windham and A.E. Yagle, “A comparative analysis of several transformations for enhancement and segmentation of magnetic resonance image scene sequences,” IEEE Trans. Medical Imaging, vol. 11, no. 3, pp. 302-316, 1992.
[9] H. Soltantian-Zadeh, R. Saigal, A.M. Haggar, J.P. Windham, A.E. Yagle and D.C. Hearshen, “Optimization of MRI protocola and pulse sequence parameters for eigneimage filtering,” IEEE Trans. Medical Imaging, vol. 13, no. 1, pp. 161-175, March 1994.
[10] H. Soltanian-Zadeh, J.P. Windham and D.J. Peck, “Optimal linear transformation for MRI feature extraction,” IEEE Transaction on Medical Imaging, vol. 15, no. 6, pp. 749-767, 1996.
[11] Wilburn E. Reddick, John O. Glass, Edwin N. Cook, T. David Elkin, and Russell J. Deaton, “Automated segmentation and classification of multispectral magnetic Resonance images of brain using artificial neural networks,” IEEE Transactions on Medical Imaging, vol. 16, no. 6, pp. 911-918, December 1997.
[12] Javad 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.
[13] J.S. Lin, R.M. Chen, Y.M. Huang, “Medical image sementation using field annealing network, “ IEEE international Conference on Image Processing, vol. 2, pp. 885-858, October 1997.
[14] J.S. Lin, K.S. Cheng, C.W. Mao, “Multispectral magnetic resonance images segmentation using fuzzy Hopfield neural network,” Interational Journal of Biomedical Computing, pp. 205-214, August 1996.
[15] J.S. Lin, K.S. Cheng, C.W. Mao, “Modified Hopfied neural network with fuzzy c-means technique for multispectral MR image segmentation,” IEEE International Conference on Image Processing, vol. 1, pp. 327-330, September 1996.
[16] M.C. Clark, L.O. Hall, D.B. Goldgof, L.P. Clarke, R.P. Velthuizen and M.S. Sibiger, “MRI segmentation using Fuzzy clustering techniques,” IEEE Engineering in Medicinee and Biology, pp. 730-742, Nov./Dec. 1994.
[17] M. Stella Atkins and Blair T. Mackiewich, “Fully automatic segmentation of brain in MRI,” IEEE Transactions on Medical Imaging, vol. 17, no. 1, pp. 98-107, February 1998.
[18] Matthew C. Clark, Lawrence O. Hall, Dmitry B. Goldgof, Robert Velthuizen, F. Reed Murtagh, and S. Silbiger, “Automatic tumor segmentation using knowledge-based techniques,” IEEE Transactions on Medical Imaging, vol. 17, no. 2, pp. 187-201, April 1998.
[19] C.-M. Wang, S-T. Yang, P.-C. Chung, C.S. Lo, C.-I Chang, C.C. Chen, C.-W. Yang and C.H. Wen, "Orthogonal subspace projection-based approaches to classification of MR image sequences," Computerized Medical Imaging and Graphics.
[20] J.C. Harsanyi, Detection and Classification of Subpixel Spectral Signatures in Hyperspectral Image Sequences, Ph.D. dissertation, Department of Electrical Engineering, University of Maryland Baltimore County, Baltimore, MD, 1993..
[21] J. Harsanyi and C.-I Chang, "Hyperspectral image classification and dimensionality reduction: an orthogonal subspace projection approach," IEEE Trans. on Geoscience and Remote Sensing, vol. 32, pp. 779-785, 1994.
[22] C.-I Chang, T.-L.E. Sun and M.L.G. Althouse, "An unsupervised interference rejection approach to target detection and classification for hyperspectral imagery," Optical Engineering, vol. 37, pp. 735-743, 1998.
[23] C.-I Chang, X. Zhao, M.L.G. Althouse and J.-J. Pan, "Least squares subspace projection approach to mixed pixel classification in hyperspectral images," IEEE Trans. on Geoscience and Remote Sensing, vol. 36, pp. 898-912, 1998.
[24] H. Ren and C.-I Chang, "A generalized orthogonal subspace projection approach to unsupervised multispectral image classification," IEEE Trans. on Geoscience and Remote Sensing, vol. 38, pp. 2515-2528, 2000.
[25] H. Soltanian-Zadeh, J.P. Windham, D.J. Peck, "Optimal linear transformation for MRI feature extraction," IEEE Trans. on Medical Imaging, vol. 15, pp. 749-767, 1996.
[26] Q. Du and C.-I Chang, "A linear constrained distance-based discriminant analysis for hyperspectral image classification," Pattern Recognition, vol. 34, pp. 361-373, 2001.
[27] C.-I Chang, "Further results on relationship between spectral unmixing and subspace projection," IEEE Trans. on Geoscience and Remote Sensing, vol. 36, pp 1030-1032, 1998.
[28] H. Stark and J.W. Woods, Probability and Random Processes with Applications to Signal Processing, 3rd ed., Prentice-Hall, NJ (2001)
[29] R.O. Duda and P.E. Hart, Pattern Classification and Scene Analysis, John Wiley and Sons, NY (1973).
[30] C.-I Chang and H. Ren, "An experiment-based quantitative and comparative analysis of hyperspectral target detection and image classification algorithms," IEEE Trans on Geoscience and Remote Sensing, vol. 38, pp. 1044-1063, 2000.
[31] C.E. Metz, "ROC methodology in radiological imaging," Invest. Radiology, 21, 720-723, (1986).
[32] H. Suzuki and J. Toriwaki, “Automatic segmentation of head MRI images by knowledge guided thresholding,” Computerized Med. Imaging and Graphics, vol. 15, pp. 233-240, 1991.
[33] K. Fugunaga, Statistical Pattern Recognition, 2nd. ed., Academic Press, NY (1990).
[34] D.H. Kil and F.B. Shin, Pattern Recognition and Prediction with Applications to Signal Characterization, American Institute of Physics, Woodbury, NY, 1996.
[35] Schwengredt RA. Remote Sensing: Models and Methods for Image Processing. 2nd ed., New York: Academic Press, pp. 470-471, 1997.
[36] Adams JB, M. O. Smith, “Spectral mixture modeling: a new analysis of rock and soil types at the Viking Lander 1 site,” J. Geophys. Res., pp. 8098-8112, 1986.
[37] Shimabukuro YE, Smith JA. “Least squares mixing models to generate fraction images derived from multispectral data,” IEEE Trans. Geosci. Remote Sensing, vol. 29, pp. 16-20, 1991.
[38] Adams JB, M. O. Smith, Gillespie AR, “Image spectrocopy: interpretation based on spectralmixture analysis,” In Pieters CM, P.A. Englert PA, editors. Remote Geochemical Analysis: Elemental and Mineralogical Composition. Cambridge University Press, pp. 145-166, 1993.
[39] Epp SS. Discrete Mathematics with Applications, 2nd ed. Brooks/Cole Publishing Company, 1995.
[40] C. Brumbley, “Kalman Filtering and Subspace Approach to Multispectral and Hyperspectral Image Classification,” Ph.D. dissertation, Department of Electrical Engineering, University of Maryland Baltimore County, May 1998.
[41] Settle JJ. “On the relationship between spectral unmixing and sub-space projection,” IEEE Trans. on Geoscience Remote Sensing, vol. 34, pp. 1045-046, 1996.
[42] C.-I Chang, C. Brumbley, “A Kalman filtering approach to multispectral image classification and detection of change in signature abundance,” IEEE Trans. Geoscience and Remote Sensing, vol. 37, pp. 257-268, 1999.
[43] Sahoo, P.K., S. Soltani, A.K.C. Wong and Y.C. Chen, "A survey of thresholding techniques," Computer Vision, Graphics and Image Process. (CVGIP), vol. 41, pp. 233-260, 1988.
[44] R.A.Schowengerdt, “Techniques for Image Processing and Classification in Remote Sensing,” New York, Academic Press, 1983.
[45] Zenzo, S.D., R. Bernstein, S.D. Degloria and H.C. Kolsky, "Gaussian maximum likelihood and contextual classification algorithms for multicrop classification experiments using thematic mapper and multispectral scanner sensor data," IEEE Trans. Geoscience and Remote Sensing, vol. GE-25, pp. 815-824, November 1987.
[46] C.-I Chang, H. Ren, Q. Du, S-S. Chiang and A. Ifarraguerri, "An ROC analysis for subpixel detection," IEEE 2001 International Geoscience and Remote Sensing Symp., Sydney, Australia, July 24-28, 2001.
[47] Zhao, X., Subspace Projection Approach to Multispectral/Hyperspectral Image Classification Using Linear Mixture Modeling, Master Thesis, Department of Computer Sciences and Electrical Engineering, University of Maryland Baltimore County, MD, May 1996.
[48] Riess, E.,”Automatic monitoring of electrical parameters in the semiconductor industry based on ROC ,“ IEEE Trans on Systems, Man and Cybernetics, vol. 30, pp. 853 -857, Nov. 2000.
[49] Gifford, H.C., Wells, R.G., King, M.A., ”A comparison of human observer LROC and numerical observer ROC for tumor detection in SPECT images,” IEEE Trans on Nuclear Science, vol. 46, pp. 1032 -1037, Aug. 1999.
[50] Woods, K., Bowyer, K.W., “ Generating ROC curves for artificial neural networks,” IEEE Trans on Medical Imaging, vol. 16, pp. 329 -337, June 1997.
[51] Wazenski, M., Alexandrou, D., DeFatta, D., ”ROC evaluation of adaptive beamforming in a simulated shallow water environment,” IEEE Trans on Oceanic Engineering, vol. 21, pp. 94 -99, Jan. 1996.
[52] Hsu, J., Chelberg, D.M., Babbs, C.F., Pizlo, Z., Delp, E.J., “ Preclinical ROC studies of digital stereomammography,” IEEE Trans on Medical Imaging, vol. 14, pp. 318-327, June 1995.
[53] Papastavrou, J.D., Athans, M., “ The team ROC curve in a binary hypothesis testing environment,” IEEE Trans on Aerospace and Electronic Systems, vol. 31, pp. 96 -105, Jan. 1995.
[54] Gooley, T.A., Barrett, H.H., “ Evaluation of statistical methods of image reconstruction through ROC analysis (emission tomography),” IEEE Trans on Medical Imaging, vol. 11, June 1992.
[55] H. Suzuki and J. Toriwaki, “Automatic segmentation of head MRI images by knowledge guided thresholding,” Computerized Med. Imaging and Graphics, 15, 233-240, (1991).
[56] J.C. Harsanyi, W. Farrand and C.-I Chang, “Detection of subpixel spectral signatures in hyperspectral image sequences,” Annual Meeting, Proceedings of American Society of Photogrammetry & Remote Sensing, Reno, pp. 236-247, 1994.
[57] R.S. Resmini, M.E. Kappus, W.S. Aldrich, J.C. Harsanyi and M. Anderson, "Mineral mapping with HYperspectral Digital Imagery Collection Experiment (HYDICE) sensor data at Cuprite, Nevada, U.S.A.," Int. J. Remote Sensing, vol. 18, no. 17, pp. 1553-1570, 1997.
[58] W. Farrand and J.C. Harsanyi, "Mapping the distribution of mine tailing in the coeur d'Alene river valley, Idaho, through the use of constrained energy minimization technique," Remote Sensing of Environment, vol. 59, pp. 64-76, 1997.
[59] S. Haykin, Adaptive Filter Theory, 3rd ed., NJ: Prentice-Hall, 1996.
[60] Frost III, “An algorithm for linearly constrained adaptive array processing,” Proc. of IEEE, vol. 60, pp. 926-935, 1972.
[61] B.D. Van Veen and K.M. Buckley, “Beamforming: a versatile approach to spatial filtering,” IEEE ASSP Magazine, pp. 4-24, April 1988.
[62] R. Schalkoff, Pattern Recognition: Statistical, Structural and Neural Approaches, John Wiley and Sons, 1992.