本篇論文主要目的是建構一套在無校準(un-calibration)環境下透過二維X光影像AP View(前後照)和Lateral View(側面照)重建三維膝關節模型，對於健保給付日愈吃緊的情況下，可用最少的資源得到最大的效益。

　　二維X光影像可利用Canny edge detector、邊界篩選、邊界搜尋演算法…等適當的前處理，並將前處理影像套用Active Shape Model (ASM)訓練完成之形狀模型，對二維X光影像分割。運用Powell的方法，調整ASM的形狀參數(shape parameter)與位置參數(position parameter)。並利用前後照分割結果自動計算膝關節幾何參數(角度與距離)。希望以客觀的方式以本系統得出之數據輔助醫師診斷病人病況，實際助於臨床上的應用。
　
　　三維通用模型的建立是使用電腦斷層掃描立體資訊利用Marching Cube重建三維膝關節模型，用此模型做為三維空間的參考基準，再將二維X光影像關節面控制點與三維投影影像關節面控制點用Iterative Clostest Point (ICP)對位並尋找控制點對應關係。用此對應關係，套用至Radial Basis Function (RBF)，將三維通用模型變形，完成二維X光影像重建三維膝關節模型之目的。

　　本系統經過實驗評估，於二維X光影像分割平均誤差在6 pixels以下；三維模型關節面平均誤差在2 pixels以下；三維模型重建誤差則僅有約5 voxels。證明本系統不論於二維X光影像分割或三維模型變形均有不錯之效果。

　The purpose of this thesis is to reconstruct the three-dimensional knee joint model from un-calibrated two-dimensional X-ray images (AP view and Lateral view). The proposed system can be useful in clinical diagnoses and helpful in economizing the use of health-insurance resources.

　In the 2D measurement and analysis of X-ray images, the images are first preprocessed using operators such as Canny edge detector, edge selection, and edge searching algorithm, …, etc.. Then, the active shape model (ASM) is adopted to segment the articular surface in the preprocessed X-Ray images. Powell’s method is used to adjust the shape and position parameters in ASM. Traditionally, assessing the knee osteoarthritis is dependent on doctors’ subjective decision. After segmenting the articular surface in the AP view image, the proposed system can automatically calculate the geometry parameters of knee-joint. These calculated parameters can provide objective evaluations and is hence helpful for clinical diagnosis.

　To reconstruct the 3D knee joint model, a model based algorithm is adopted. First, a generic knee joint model was built from CT volume data by using the marching cube algorithm. Then, control points of the ASM defined on the 2D images were registered to the projection images of the 3D generic model by using the iterative closest point (ICP) algorithm. The generic model is then deformed by using the radial basis function (RBF) to obtain the 3D knee joint model of the given 2D X-ray images.

　In the experiments, the average segmentation errors of articular surfaces are all under 6 pixels. In 3D model reconstruction, the average projection and the average reconstruction errors are under 2 pixels and 5 voxels, respectively. From the experimental results, the proposed system achieves good performances in both 2D articular surface segmentation and 3D model reconstruction. It is useful for knee osteoarthritis assessment and can also be applied to other clinical applications.

[1]S. Martinez, M. Korobkin, F. B. Fondren, L. W. Hedlund, and J. L. Goldner, “Diagnosis of patellofemoral malalignment by computed tomography,” Journal of Computer Assisted Tomography 7(6), pp.1050-1053, 1983.

[2]S. F. Schutzer, G. R. Ramsby, and J. P. Fulkerson, “The evaluation of patellofemoral pain using computerized tomographpy,” Clinical Orthopaedics and Related Research, pp. 286-293, 1986.

[3]F. G. Shellock, J. H. Mink, and J. M. Fox, “Patellofemoral joint: kinematic MR imaging to assess tracking abnormalities,” Radiology 168, pp. 551-553, 1988.

[4]F. G. Shellock, J. H. Mink, A. L. Deutsch, and J. M. Fox, “Patellar tracking abnormalities: clinical experience with kinematic MR imaging in 130 patients,” Radiology 172, pp. 799-804, 1989.

[5]F. G. Shellock, J. H. Mink, A. L. Deutsch, and T. K. Foo, “Patellofemoral joint:evaluation during active flexion with ultrafast spoiled GRASS MR imaging,” Radiology 180, pp. 581-585, 1991.

[6]F. G. Shellock, J. H. Mink, A. L. Deutsch, and T. K. Foo, “Kinematic MR imaging of the patellofemoral joint: comparison of passive positioning and active movement techniques,” Radiology 184, pp. 574-577, 1992.

[7]F. G. Shellock, J. H. Mink, A. L. Deutsch, T. K. Foo, and P. Sullenberger, “Patellofemoral joint: identification of abnormalities with active movement, ‘unloaded’ versus ‘loaded’ kinematic MR imaging techniques,” Radiology 188, pp. 575-578, 1993.

[8]J. Brossmann, C. Muhle, C. Schrder, U. H. Melchert, C. C. Bll, R. P. Spielmann, and M. Heller, “Patellar tracking patterns during active and passive knee extension with motion-triggered cine MR imaging,” Radiology 187, pp.205-212, 1993.
[9]A. H. Newberg and D. Seligson, “The patellofemoral joint: 30o, 60o, and 90o views,” Radiology 137, pp. 57-61, 1980.

[10]A. P. Davies and M. M. S. Glasgow, “Imaging in osteoarthritis: a guide to requesting plain X-rays of the degenerate knee,” The Knee, no. 7, pp. 139-143, 2000.

[11]S. A. Mazzuca, K. D. Brandt et. al., “Effect of alignment of the medical tibial plateau and X-Ray beam on apparent progression of osteoarthritis in the standing anteroposteior knee radiograph,” Arthritis & Rheumatism, vol. 44, no. 8, pp.1786-1794, 2001.

[12]David Siu, et al., “Three dimensional reconstruction of joint surface using a microcomputer,” Med. & Bio. Eng. & Comput., 24, pp. 267-274,1986.

[13]David Siu, et al., “Femur articular shape and geometry: a three-dimensional computerized analysis of the knee,” The Journal of Arthoplasty, vol. 11, no. 2, pp. 166-173, 1996.

[14]Andre B., Dansereau J., Labelle H., “Optimized vertical stereo base radiographic setup for the clinical 3D reconstruction of the human spine,” Journal of Biomechanics, vol. 27, pp. 1023-1035, 1994.

[15]Mitton, D., Landry, C., Vron, S., Skalli, W., Lavaste, F., de Guise, J.A., “A 3D reconstruction method from biplanar radiography using non stereocorresponding points and elastic deformable models,” Medical and Biological Engineering and Computing, vol. 38, pp. 133-139, March 2000.

[16]Laporte S., Skalli W., De Guise J.A., Lavaste F., Mitton D., “A Biplanar Reconstruction Method Based on 2D and 3D Contours: Application to the Distal Femur,” Computer Methods in Biomechanics & Biomedical Engineering, Vol. 6, Issue 1, pp. 1-6, Feb. 2003.

[17]Bras, Laporte, Bousson, Mitton, Guise, Laredo and Skalli, ”Personalized 3D reconstruction of proximal femur from low-dose digital biplanar radiographs,” International Congress Series, 1256, pp. 214-219, 2003.

[18]Pomero et al., “Fast semiautomatic stereoradiographic reconstruction of scoliotic spines using multi-scale image processing and statistical geometric models,” International Congress Series, 1256, pp. 207-213, 2003.

[19]Laporte et al., “3D reconstruction of the rib cage from biplanar X-rays,” International Congress Series, 1268, pp. 1275, 2004.

[20]Hugo Embrechts, Dirk Roose, ”A Parallel Distance Transformation Algorithm,” Distributed Memory Computing Conference, 1991. Proceedings., The Sixth April 28-May 1, pp. 479 – 482, 1991.

[21]Cootes, T.F., Taylor C, J., “Active shape models – smart snakes”, Proceedings British Machine Vision Conference, Springer, Berlin, pp. 266-275, 1992.

[22]N. Otsu, “A threshold selection method from gray-level histogram,” IEEE Transaction on System, Man, and Cybernetics, Vol. 9, No. 1, pp. 62-66, January 1979.

[23]Lorensen, W.E. and Cline, H.E., “Marching Cubes: a high resolution 3D surface reconstruction algorithm,” Computer Graphics (Proc. of SIGGRAPH), Vol. 21, No. 4, pp. 163-169, 1987.

[24]Paul J. Besl, "A method for registration of 3-D shapes," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 14, no. 2, pp. 239-256, February 1992.

[25]Milan Sonka, Vaclav Hlavac, Roger Boyle, “Image Processing, Analysis, and Machine Vision,” Second edition, PWS publishing, 1998.

[26]Adam Greenspan, MD, “Orthopedic radiology: a practical approach,” Second edition, Gower Medical publishing, 1992.

[27]Singiresu S. Rao, “Engineering Optimization: Theory and Practice,” Third edition, Wiley-Interscience, 1996.

[28]Simon Haykin, “Neural Networks: a comprehensive foundation,” Second edition, Prentice Hall publishing, 1999.

[29]鄭慶明, “實用影像診斷學,” 俊傑書局發行, 2001.

[30]陳博光, “骨頭毛病多,” 元氣齊出版社發行, 1995.

[31]鄭慶明, “電腦斷層掃描攝影大全,” 合記圖書出版社發行, 1990.