由於銀髮族時代來臨，老年人口變多，脊椎骨病變的案例已成為醫療常見問題，使得脊椎方面的手術病例增加，而脊椎手術中最常見步驟即是骨釘植入融合手術。骨釘植入融合手術是永久連接兩節或多節脊椎骨，以減輕骨節之間的運動造成的疼痛。此手術仰賴專業醫生的技巧與經驗，且骨釘的下竿點與路徑對於手術成功與否十分重要。因此本研究以建置一套手術訓練系統為目標，提供新手醫生練習手術與熟悉操作步驟的情境系統。此系統需要脊椎骨的視覺模型，於是在此提出了一個基於重新分類法的脊椎骨影像分割，由脊椎的臨床電腦斷層影像提供的高對比骨骼影像資訊來進行脊椎模型的重建。另外，也結合由影像對位建構出的核磁共振三維影像，提供影像對比較明顯的軟組織 (血管與神經) 資訊與結構，以重建出骨頭週遭的軟組織模型。然後將骨骼模型對位至核磁共振影像，結合軟組織模型共同組合成一個完整的脊椎影像模型，以為模擬手術操作之用。同時還設計了基本的手術操作情境，利用上述建立的脊椎模型與手術器具模型，在電腦螢幕窗口上模擬手術。對器具於三維空間中的座標轉換以及與其他組織模型的碰觸偵測，給予類似實際臨床上的模擬。系統具有數項小功能的設計，包括碰觸軟組織的亮燈警示，與手術完成後脊椎骨半透明化顯示骨釘位置，與操作過程之紀錄重播等，以幫助手術醫生了解目前系統的操作情形。
關鍵字：脊椎骨、電腦斷層、核磁共振、影像對位、模型對位、情境模擬

Due to the fast increase of aging population, the need of spinal surgery increases as well. The rapidly increasing cases of spinal surgery become a major concern in aging care recently. One of the most popular treatments of spinal surgery is the spinal fusion surgery. Spinal fusion is a surgery to permanently connect two or more vertebras in order to reducing the pain by eliminating motion between them. This surgery relies on the professional surgeon’s technique and experience; the inserting position and the drilling path of pedicle screws play a critical role and affect the success of surgery. Therefore, in this research, a surgical training system is designed and implemented to provide novice surgeons with a simulation system which can be used for practicing pedicle screw implantation and getting familiar with surgical procedures. In this system, a bone segmentation method based on image reclassification was first designed to construct the spine model from computer tomography (CT) images which provide high contrast information of bone structure. Moreover, the magnetic resonance (MR) images from different scanning modes were combined after image registration to form a more complete MR image volume which provides better contrast information of soft tissues in order to reconstruct the soft organ models, such as artery vein or nerves, around the spine. By applying bone model registration to the MR images, the soft organ models were integrated with the bone model to constitute a complete spine image model which was then used in the surgical simulation system. With the complete spine image model and the designed surgical tool models, the scenario of surgical operations was designed to simulate the spine surgery on the computer screen. The positioning and path of tool models and the detection of tool cutting with other organs were designed and implemented to simulate the operations of real clinical surgery in 3D. Some additional system functions, including warning light detection when touching soft tissue model, making spine model semi-transparent to display the position of pedicle screws, and after surgery replay of recorded surgical procedure, were also implemented in order to help surgeons better understand the situations of simulated operations after the use of training system.
Keyword: spine, computed tomography, magnetic resonance, image registration, model registration, scenario simulation

[1]http://education.yahoo.com/reference/gray/subjects/subject/25
[2]http://www.kidport.com/reflib/science/humanbody/skeletalsystem/Vertebrae.htm
[3]http://www.energycenter.com/grav_f/anatomy.html
[4]Manbachi, A., Cobbold, R.S.C. and Ginsberg, H.J., "Guided pedicle screw insertion: techniques and training," Spine J 14, 2014, pp.165-79.
[5]Hamza-Lup FG, Bogdan CM, Popovici DM, Costea OD. "A survey of visuo-haptic simulation in surgical training. "eLmL 2011. In: The third international conference on mobile, hybrid, and on-line learning; 2011, pp.57–62.
[6]Min-Tzu Chu, (2013) “Development of image guided surgical training system: Model, calibration and scenario”
[7]Naegel, B. "Using mathematical morphology for the anatomical labeling of vertebrae from 3D CT-scan images." Computerized Medical Imaging and Graphics, 31(3), 2007, pp.141-156.
[8]Ma, Lu, "Hierarchical segmentation and identification of thoracic vertebra using learning-based edge detection and coarse-to-fine deformable model", Computer Vision and Image Understanding 2013, pp.1072-1083.
[9]Kim, Y., & Kim, D. (2009). A fully automatic vertebra segmentation method using 3D deformable fences. Computerized Medical Imaging and Graphics, 33(5), 343-352.
[10]Klinder T, Ostermann J, Ehm M, Franz A, Kneser R, Lorenz C. Automated model-based vertebra detection, identification, and segmentation in CT images. Med Image Anal. 2009; 13: 471–82
[11]C. Lorenz, N. Krahnstöver "Generation of point-based 3D statistical shape models for anatomical objects "Computer Vision Image Understand, 77,2000, pp.175–191.
[12]Y. Cheng, S. Zhou, Y. Wang, C. Guo, J. Bai, S. Tamura "Automatic segmentation technique for acetabulum and femoral head in CT images". Pattern Recogn, 46 (11) 2013, pp.2969–2984.
[13]Dalca, A., et. al. "Segmentation of nerve bundles and ganglia in spine MRI using particle filters", 2011, pp. 537-545. Springer Berlin Heidelberg.
[14]Gerig, G., Kubler, O., Kikinis, R., & Jolesz, F. A. (1992). Nonlinear anisotropic filtering of MRI data. Medical Imaging, IEEE Transactions on, 11(2), pp.221-232.
[15]Hae Yong Kim “Gradient Histogram-Based Anisotropic Diffusion”, Universidad de Sato Paulo, Escola Politécnica, Dept. Eng. System’s Electronics, Av. Prof. Luciano Gualberto, tr. 3, 158, 05508- 900
[16]W. K. Pratt, Digital Image Processing 4th Edition, John Wiley & Sons, Inc., Los Altos, California, 2007.
[17]Brent, Richard P. (1973)." Section 7.3: Powell's algorithm." Algorithms for minimization without derivatives. Englewood Cliffs, N.J.: Prentice-Hall.