用於電腦輔助骨科手術之導航系統開發

Abstract

機器人輔助手術中，有三個重要元素：模型、機器人與模型-機器人座標系統轉換關係。模型係手術標的之資訊總和，根據手術應用，提供標的外表形狀及/或力學資訊。機器人則由醫師操作以進行手術，包含傳統器械及主動式引導或保護器具。兩者的整合可以透過之間座標轉換達成，這也是輔助手術系統中最重要的一部分，直接影響手術的精度。輔助系統的目的，在於手術提供即時之模型與機器人資訊，以提高非侵入性、降低手術誤差並最佳化手術結果；但不能因導入輔助系統而增加過多的手術時間、手術繁複度。故本研究的重點在於對不同手術應用找出一套更有效率且較少侵入性的模型-機器人轉換方案。 本論文以兩個普遍的骨科手術應用，探討兩種導航系統之開發：全膝關節置換手術，係將受損膝關節內的軟骨、韌帶、骨頭表面切除，植入人工關節，以重新提供膝關節之運動功能，即使手術失敗亦不會造成生命危險。在第二章中，藉由使用蒐集到關節運動數據計算髖關節中心與膝關節旋轉軸，對於全膝關節置換提出一更具效率的模型建立方法。由於作為模型建立基準之標記，需固定於手術目標周遭之骨頭中，需額外之切口，故於第三章中，透過分析皮膚固定標記取代骨骼固定標記之可行性，嘗試降低模型建立的侵入性。 而椎弓釘固定是脊椎手術常見的植入物固定方式。骨釘沿著椎弓進入椎體內完成固定。由於骨釘和椎弓尺寸接近，且結構複雜、緊臨神經，植入角度、深度稍有偏差即可能傷到神經，帶來不可回復之傷害，具有高度風險性。其導引模型通常需要使用病患自身電腦斷層影像作為基準，醫師需要在術前花費額外之時間於模型上進行手術規劃、模擬。在開放式脊椎手術(第四章)中，因為醫師可以直接取得手術目標之表面拓樸，與術前影像疊合，故探討不同取點區域或區域組合對疊合精度之影響。在微創手術(第五章)中，無法直接接觸目標表面，需通過術中X光影像完成手術。為了疊合病患和導引模型，需將術前之CT影像進行投影，與實際術中影像進行比對，以找出最可能之姿態。 經過兩種手術的不同施行方式可得知：針對不同適應症，手術模型應有所調整，連帶著工作流程亦需有所變動。膝關節手術講究效率，如何減少時間花費為開發重點，可藉由直接於術中建立模型，減少醫師負擔。而脊椎手術具有高風險性，需要以電腦斷層影像作為準確之手術依據，術前之規畫與模擬亦不可缺少。不論何種手術與工作方式，模型與手術標的疊合直接影響手術精準度，善用解剖特徵找出模型-機器人座標轉換關係，更是輔助系統於開發過程中應考慮之重點項目。

In computer-assisted or robot-assisted surgery, there are three important components of the assisted system. That is model, robot and transformation between model and robot coordinates: Model is a combination of information of surgical target, such as the appearance model and/or mechanical properties of target according to the surgical indications. Surgeon manipulates robots to operate the surgery. With the presence of power, the robot is classified as passive instruments and active robots for guiding or protection. The relationship between coordinates of model and robot is used integrate the two components which is the essential part of computer-assisted system affecting the accuracy directly. Computer-assisted navigation has a role in some orthopedic procedures. It offers the potential to decrease intraoperative errors, optimize the surgical result and make the surgery less invasively by providing real-time feedback. But the introduction of assisted system must not increase the operation time and procedures. Therefore, the purpose of this study is to find a protocol to build the transformation between model and robot for different orthopedic surgical application with more efficiency and less invasiveness. Based on two popular applications of orthopedic surgery, this study explores the development of navigation systems in two extremes: Total knee replacement is a set of surgical procedures cutting away the damaged bone and cartilage in the end of the femur and tibia to replace the diseased or damaged knee joint by an artificial one. For the purpose to restore knee function, surgery failure is not fatal. In Chapter 2, a model-building method with more efficiency for total knee replacement is proposed by collecting the kinematic data and calculating the center of hip joint and the rotation axis of knee joint. As references of navigation surgery, it needs fiducial markers rigid fixed on bones near to the target through an extra incision. For decreasing the invasiveness, a feasible analysis for replacement of traditional skeleton-pined markers with skin-attached ones was employed in Chapter 3. Fixing implants through pedicles of vertebra by screws is common in spinal surgery. Due to the similar sizes of screw and pedicle and the complex anatomical structure, it exists high risks that nerve damage caused by the error of operation and the damage may be irreversible and fatal. Surgeons have to spend additional time for pre-operative planning and simulation on the model developed from the patient’s own CT image. In the open spine surgery (Chapter 4), surgeon can access the surgical target directly. A collection of 3D points from vertebral surface of the patient is matched to the specific image-based model by iterative closest point (ICP) algorithm. A discussion about the registration accuracy affected by point-sampling from different region or region combinations was raised. On the other hand, surgeon cannot access the surgical target directly and carries out the operation with a mobile image intensifier system (C-arm) in the minimally invasive surgery (Chapter 5). For purpose of registration, set of digitally reconstructed radiographs (DRR) is generated to simulate the image taken intra-operatively by C-arm. A similarity function is used to evaluate the posture of CT image with the highest possibility In conclusion, models differ from the indications of surgeries with different requirement of accuracy, so do the workflows. The model for application of TKA can be simplified as geometric model and be constructed intra-operatively to reduce time consumption. In fixation of pedicle screw, a more precise model based on CT image is needed, in addition to pre-operative planning and simulation. Both the two applications, registration process is the most important for integrating model and robot and affects the surgical accuracy directly Make a good use of anatomical feature for registration may be worthy of attention for development of navigation system.