目前最常使用的長骨骨折固定方式是在骨髓腔植入骨髓內釘，但由於骨髓內釘在植入過程中會變形，難以固定式鑽孔導引器輔助鑽孔，因此目前多依賴大量的C-arm X光影像輔助醫師進行鑽孔，所累積的輻射劑量對醫護人員以及病患健康都存著潛在的傷害。因此發展一可使用少量X光影像, 成本又相對低的骨髓內釘鑽孔導引系統是有其需要性。 本研究以C-arm影像結合高效率透視N點攝影機姿態估計方法(EPnP)進行定位板相對於C-arm影像坐標系的空間姿態估計，並拍攝兩組不同角度之骨髓內釘C-arm影像，利用雙角度攝影空間定位技術計算出骨髓內釘長軸方向及遠端固定孔中心位置。接著依據C-arm拍攝時的投影模式，將骨髓內釘模型置放在上述計算得到的骨髓內釘位置，並繞自軸旋轉，旋轉過程中將骨髓內釘模型的固定孔投影影像與實際骨髓內釘固定孔影像進行相似度比較，其中相似度最高的骨髓內釘模型的固定孔軸向即是真實骨髓內釘固定孔軸向，再利用雙角度攝影空間定位技術計算出相對於定位板坐標系的該骨髓內釘固定孔空間軸向。由於鑽孔導引器與定位板之間的座標轉換是已知的設計值，因此可經由座標轉換，調整鑽孔導引器的導槽方位對準骨髓內釘固定孔軸向及中心位置，即可協助醫師進行鑽孔導引。 實驗在骨髓內釘固定孔兩側各拍攝3張與4張影像，共組合出12組案例，實驗結果顯示其中10組可導引鑽頭進入骨髓內釘固定孔，另外兩組則由於C-arm影像解析度以及影像雜訊處理的不夠好造成導引失敗。 ;Currently intramedullary nailing is a common operation for long bone fracture reconstruction. Distal locking of intramedullary nails is difficult due to the nail deformation after implantation. Therefore, orthopedic surgeons take tens C-arm images to verify the correction of drilling procedure. The accumulation of mass radiation dosage is harmful to the health of the patient and medical persons. Thus, development of a low-cost and low-radiation navigation system for intramedullary nailing is expected. In this research, the integration of C-arm images with efficient perspective-n- point camera pose estimation (EPnP) method is applied to calculate the spatial position and orientation of the positioning plate relative to the C-arm image frame. Two C-arm images taken in different positions are used to determine the axial direction of the intramedullary nail and the center position of fixation holes by Bi-plane algorithm. Then, according to the real C-arm X-ray projection model, an identical computer model is created to simulate the C-arm X-ray projection. The intramedullary nail is rotated about its axis sequentially and the images of fixation holes are projected simultaneously. Similarity measures of the projection images of fixation holes generated by computer simulation and the real C-arm system are calculated. The axial directions of fixation holes of the simulated intramedullary nail with largest similarity measure value are those of the real intramedullary nail. Then, the guide slot of the drill guide device will be adjusted to have identical directions as the fixation holes. In the experiment, 3 and 4 C-arm images are taken from both sides of the fixation holes respectively. They are grouped to generate 12 experimental cases. The experimental results show that there are 10 cases which the drill bit can be guided to pass through the fixation holes. The other 2 cases are failed to do it because of image processing and segmentation of steel balls of the positioning plate and fixation holes of the intramedullary nail are not good enough.
