本論文提出一個膝關節十字韌帶自動化診斷系統，此系統針對側照膝關節X光影像作處理，從影像上分割出在臨床分析上有意義的區域並定義出特徵點，最後利用特徵點計算用以診斷的特徵值；而除了當前醫學上使用的特徵值外，我們亦提出其他可用以診斷的特徵值。在當前十字韌帶完整度的診斷上，皆為醫師以手動的方式量測，並在MR影像上做判斷，因此其診斷的準確率會因醫師的不同而產生差異。
在側照膝關節X光影像上，除了有對比度不佳的問題之外，其關節部位的結構也相當複雜。因此在分割的方法上，我們使用以模型為基礎的分割法，對影像上有意義的區域建立其在外觀以及形狀上的模型，並將影像中與模型最相似的區域分割出來。而為因應結構上較為複雜的兩重疊骨髁區域，我們建立雙骨髁的模型，在考慮到兩骨髁的對應關係下同時將兩骨髁分割出來。
在最後的驗證上，我們計算所測量出的脛骨位移在十字韌帶臨床診斷上的準確率，並對一些影響準確率的因素作分析。

An automatic image analysis system for the diagnosis of knee cruciate ligament is presented. This system segments objects which are used to define some feature points and parameters for cruciate ligament evaluation. Other than some conventional parameters given in literature, we also design new parameters which can be used for cruciate ligament diagnosis.
The current cruciate ligament diagnosis is commonly done by doctor’s manual measurement based on the Magnetic Resonance images. Therefore, the diagnostic accuracy varies from person to person. Because of the poor contrast and the complex structure of the knee, we apply a model-based segmentation algorithm to segment out objects of interest from the lateral view of knee radiograph. The appearance model and the shape model of objects of interest in the image are established and used for segmentation. By using the proposed model-based algorithm, the contour which best matches with the model can be successfully segmented. For the complicated femoral condyles, we use a dual condyles model which simultaneously segments two condyles using the relationship between the condyles.
In the experiments, the diagnostic accuracy of each measurement is promising. Furthermore, several factors that affect the reliability of measurements are also analyzed.

[1] P. A. Torzilli, R. L. Greenberg, R. W. Hood, H. Pavlov, J. N. Insall, “Measurement of anterior-posterior motion of the knee in injured patients using a biomechanical stress technique,” The Journal of Bone and Joint Surgery, vol. 66-A, pp.1438-1442, 1984.
[2] P. Wirz, P. von Stokar, R. P. Jakob. “The effect of knee position on the reproducibility of measurements taken from stress films: a comparison of four measurement methods,” Knee Surg Sports Traumatol Arthrosc, vol. 8, pp.143-148, 2000.
[3] Y. S. Lee, S. H. Han, J. Jo, K. S. Kwak, K. W. Nha, J. H. Kim, “Comparison of 5 Different Methods for Measuring Stress Radiographs to Improve Reproducibility During the Evaluation of Knee Instability,” AJSM PreView, 2011.
[4] H. C. Tsai, “Computer Image Assessment and Analysis for Knee X-ray Image,” 國立成功大學資訊工程所碩士論文, 2007.
[5] D. Seghers, D. Loeckx, F. Maes, D. Vandermeulen, P. Suetens, “Minimal shape and intensity cost path segmentation,” IEEE Trans Med Imaging 26, pp. 1115-1129, 2007.
[6] T. F. Cootes, C. J. Taylor, D. H. Cooper, J. Graham, “Active shape models — their training and application. Comput,” Vis. Image Underst. 61, 38–59, 1995.
[7] I. Sobel, “Camera models and machine perception,” PhD Thesis, Stanford University, 1970.
[8] S. Seifert, I. Waechter, G. Schmelzle, R. Dillmann, “A knowledge-based approach to soft tissue reconstruction of the cervical spine,”IEEE Trans. Med. Imag, vol. 28, no. 4, pp. 494-507, 2009.
[9] http://lithosphere.lithium.com/t5/Lithium-s-View/The-True-Marketing-Power-of-
Facebook-Sociology-Perspective/ba-p/18520
[10] http://en.wikipedia.org/wiki/Ellipse
[11] P. C. Mahalanobis, "On the generalised distance in statistics," Proceedings of the National Institute of Sciences of India, vol. 2, pp.49-55, 1936.
[12] R. C. Gpmzalez, R. E. Woods, “Digital Image Processing,” 2002.