C-arm影像輔助手術導引系統可使用兩張C-arm影像的條件下精確得知骨骼與手術器械的相對位置，具有減低輻射線的暴露量、提升手術品質、縮短新進醫師學習曲線等優點。因此本研究以實驗室既有之C-arm導引系統為基礎針對脊椎手術臨床需求進行分析與系統改良。C-arm導引系統包含影像處理、座標轉換與X光投影模式建構等。系統透過影像處理方法自動偵測C-arm校正器上大、中、小型標記在影像上的座標以供影像校正與座標對正使用。本研究改良大、小型標記之影像辨識方法，提出高通濾波法萃取大鋼珠輪廓、自適應門檻值法分割小鋼珠影像，針對臨床所取得之98張C-arm影像進行測試。結果顯示大型標記之整體辨識率由原先的62%提升至98%，小型標記辨識率75%以上的影像則由35張增加至80張，而且沒有辨識率低於60%之影像。又C-arm影像拍攝與擷取過程，病人的呼吸會造成目標病灶與C-arm影像校正器相對位置的改變。正確的X光投影模式必須在C-arm設備拍攝影像瞬間紀錄病患身上與C-arm影像校正器上DRF的座標值。本文提出一種手術導引系統之X光影像同步校正方法，該方法包含四個步驟：1.量測影像成像與輸出時間差、2.取得座標轉換關係、3.測量影像相似性、4.重建影像與座標轉換關係。電腦端偵測影像更新時依事先量測的輸出延遲時間取拍攝瞬間的座標關係，解決影像拍攝與擷取過程病人呼吸造成投影模式錯誤的問題。C-arm guidance system is developed by integrating image processing, coordinate transformation and projection model estimation. Only two C-arm images are required to compute the orientations and positions of the targets. The locations of surgical instruments are also real-time shown on the computer displayed C-arm images, which enable the surgeon to move surgical tools accurately and safely to the target for following treatment. It also reduces radiation exposure during the operation.The system automatically detects coordinates of the marker on the C-arm image by image processing techniques. Those markers are used in image distortion and registration. If the number of being recognized markers could not exceed the system requirements, an additional image is required. Therefore, the level of the recognition rate will affect the ease of use of the guidance system.To improve the successfully recognition rate we analysis 98 C-arm images obtained from clinical try. The overall recognition rate of large marker is increased from 62% to 98%. As for small marker, the quantity of images which recognition rate above 75% is increased from 35 to 80. And recognition rate of all the images are higher than 60%. Due to patient’s breathing, the relative position between C-arm image calibrator and patient is changed during the process of image capture. Using non-synchronized images and coordinate data will cause an incorrect estimation of projection model. We develop a kind of X-ray images synchronized registration method used in surgical guidance system, the method includes four steps: 1. Estimate the latency of fluoroscopic imaging. 2. Continuously obtain relative position between C-arm image calibrator and patient. 3. Measuring image similarity. 4. Reconstruct the relation between images and coordinate transformation. When the C-arm image is refreshed, software would automatically choose a transformation that has a timestamp prior to pre-measured latency. It solves the problem that image and coordinate are non-synchronized.
