傳統的腦部手術存在傷口大與對腦部組織傷害大的缺點，而微創內視鏡手術具有傷口小、病患術後回復速度快與對腦組織傷害小的優點，因此在近十年來蓬勃發展。當醫師在執行腦部微創內視鏡手術時，由於腦部的血管與神經分布複雜，且進入病患腦部之後，醫師只能透過內視鏡影像、CT/MRI影像與自身的解剖知識和臨床經驗來判斷目前內視鏡影像區域後面可能有什麼重要組織，因此在手術過程中必須小心地操作手術器械，以免傷到重要的腦動脈血管或腦神經組織，故術前規劃出一條盡量避開腦部重要組織的手術路徑，且提供一種直觀的方式讓醫師可以確認手術路徑姿態，並在手術的過程中，可以提前告知醫師被內視鏡影像區域後面有什麼組織，讓醫師有更多資訊做判斷，以提高手術安全性。 本研究以實驗室既有的CT影像導航系統為基礎，透過多路復用輸入輸出模型的套用與資料傳遞機制的建立，將實驗室既有的AR系統整合到其中，利用AR眼鏡的虛實對位，提供醫師一個直觀且3D的方式來觀察規劃的手術路徑在實際病人身上的姿態，並新增多台AR眼鏡同時使用的功能，使AR系統可以應用於協作或教學的場合。本研究改良了原CT影像導航系統的手術路徑規劃方式，讓醫師可以在點選結束點後，利用系統介面上的拉桿來調整手術路徑的方向；又透過相機校正、手眼校正與虛實影像疊合的方式，新增了內視鏡導航功能與呈現內視鏡影像區域背後血管組織的預警系統，提供醫師更多資訊做手術安全判斷。 本研究完成了內視鏡相機校正驗證實驗與內視鏡手眼校正驗證實驗，內視鏡相機校正實驗的特徵點平均位置誤差為0.75±0.49pixel；內視鏡手眼校正驗證實驗的平均特徵點位置誤差為0.41±0.13mm，在導航系統端的準確度(Accuracy)良好的狀況之下，實驗結果是符合臨床可容許誤差(2mm)。 ;Comparing with conventional brain surgery, minimally invasive endoscopic surgery has the advantages of small wounds, rapid recovery of patients after surgery, and less damage to brain tissues. Therefore, minimally invasive endoscopic surgery has been booming in the past decade. When surgeons perform the minimally invasive brain endoscopic surgery, they can only use the endoscopic and CT/MRI images with their own anatomic knowledge and clinical experience to guess the existence of brain vessels or nerves behind the endoscopic image. Therefore, the surgeons need to be very careful intraoperatively to avoid damaging cerebral arteries or nerves, causing cerebral edema and intracranial hemorrhage. This research develops an endoscopic surgical navigation system, by modifying the path planning method and integrating the AR system into the existing CT image based navigation system to provide an easy and intuitive way for planning a safe surgical path that can avoid important brain tissues such as cerebral arteries and nerves. This research also adds new path planning functions and virtual endoscopic images of arteries onto the current endoscopic image so that the surgeons realize the existence of arteries behind the current endoscopic image. Two experiments of this research can were completed. The first experiment is to measure the accuracy of camera calibration, and the results show that the average distance error of feature points is0.75±0.49pixel. The second experiment is to measure the accuracy of hand-eye calibration. The results show that the average distance error is 0.41±0.13mm. Under the good condition of the accuracy of the navigation system, then the result of this experiment is lower than the clinical requirement range (2mm).
