影像辨識自動註冊應用於機械式手術導航系統

Abstract

隨著電腦運算速度與術中影像設備技術的提升，醫學影像也臻於成熟，衍生多樣應用可能性，使得微創手術概念得以實踐，而醫學影像導航系統在微創手術中扮演不可或缺的角色，然而在臨床場域中導航系統仍不算普及，可能原因有操作流程繁複且成本高昂等，因此本研究以自動化且使用操作簡易為任務標的，希冀能順利融入現行手術流程中，提高醫師採用意願。 在過去本實驗室研究成果中，成功建立結合3D 掃描技術之機械式導航系統，實現無標記定位，由患者身體表面特徵建立患者位置與影像之座標關係，並以顱部假體進行導航誤差分析，但此系統於使用流程中仍須手動操作，操作機械手臂進行近距離的3D掃描量測定位，因此流程複雜度提升，也可能因不慎操作導致器械勿觸患部產生感染風險，並掃描定位完畢後，必須人工標示特徵點以進行座標系註冊，流程複雜且費時，並且無法適用於背部無明顯特徵點的脊椎手術中。 本研究精進機械式導航系統流程自動化，沿襲過去雷射導引機械手臂架構，結合自行開發之立體視覺系統，立體視覺系統拍攝範圍只要涵蓋定位器即可自動辨識定位點完成註冊，使導航系統更加接近臨床需求，並且整合系統軟硬體各部分以C++程式語言開發了導航規畫軟體，此一軟體具備醫學影像顯示、自動定位與註冊、穿刺路徑規劃並輸出等等功能，高度整合性與自動化技術更可推展到近年逐漸引進的術中影像設備，當患者位置變動時得以快速重建與影像之座標關係，近乎無縫搭配影像即時導航。最終針對脊椎手術進行本研究開發之導航系統誤差分析與探討，以仿脊椎模型進行測試，提出誤差產生原因以及未來改良方向。

With a leap of computer technology and medical image, various potential applications are derived. The concept of minimally invasive surgery becomes practical, and medical image navigation system plays an important role in the surgery. However, the navigation system is still not widely applied in the operating room. The complicated operation and high cost are the main factors. This study based on automation is hoped to integrate the navigation system developed into the current surgical procedures and improve the physicians’ practice. In the previous research results of our laboratory, a mechanical navigation system combined with 3D scanning technology was successfully established to realize the marker-free registration. The relationship between the patient's position and the image was established by the surface features of the patient's body. However, the system still needs to be manually operated during the navigation process. The robot needs to be operated neer the patient for 3D scanning measurement. The risk of infection may increase. Also the feature points on the patient need to be manually marked for registration. This procedure is also complicated and time consuming. Furthermore, the 3D scanning cannot be applied to spinal surgery without obvious feature points on the back. This research focuses on the process automation of a laser-guided mechanical navigation system. This navigation system is combined with the stereo vision system, which can automatically identify the marker. A development of a software including medical image display, automatic registration, puncture path planning, etc has been made. This technology can be equally applied to the intraoperative image equipment. When the patient's position changes, the relationship between the image and the patient can be quickly reconstructed, and physicians can carry out instant navigation accordingly. Finally, we use the spine model for the error analysis and discus the navigation system developed in this study. The causes of error and future improvement directions are proposed in conclusions.