多機擴增實境應用平台於臨床術中手術導引

Abstract

擴增實境(Augmented Reality, AR)是一項嶄新的技術，其允許使用者透過一個頭戴裝置來於自己的現實視野中出現虛擬的物件，其應用價值十分廣泛，尤其在醫療領域正如雨後春筍般地發展。此論文研究使用一個多機擴增實境之平台系統，並應用在聚焦超音波腦部手術以及腹部肝臟穿刺手術，使得手術流程更加簡易且快速，此外更具有展示、教學與模擬的價值。此外，為增加應用的場景之彈性及優化其效果，此論文研究提出相應的方法，最終也設計了一套實驗流程來驗證使用者量測之精準度，並量測該應用時的最大作業距離及幀率。在優化的前置實驗中，我們可以確保執行時幀率在45fps左右，以及製作出在不同應用場景時符合作業距離下且大小適當的錨點；在精準度的實驗方面，若為多用戶皆能看到錨點的情況下，平均誤差為2.7mm，若為錨點被部分遮擋的情況下，平均誤差則為7.16mm，此外也可觀察到使用者距離越遠的情況下平均誤差也會稍微變大，但因皆位於錨點的辨識距離內故誤差也相去不遠。而在不同的應用場景下，也會有相去不遠之結果，故可推出此平台系統對於HoloLens1是非常具有可行性，可以容易地經過適當的修改移植於其他的應用場景中。

Augmented Reality (AR) is a new technology that allows users to visualize virtual objects in their own real field through a headset, and its application value is very wide, especially in the medical field. This thesis research uses a multi-machine augmented reality platform system and is applied to focused ultrasound brain surgery and abdominal liver puncture surgery, making the surgical process easier and faster, and has more value for display, teaching and simulation. In addition, in order to increase the elasticity of the application scenario and optimize its effect, this paper also proposes corresponding methods, and finally designs a set of experimental processes to verify the accuracy of user measurement, and measure the maximum working distance and frame rate of the application. In the optimized pre-experiment, we can ensure that the frame rate is around 45fps when executing, and make anchor points of the appropriate size at the working distance for different application scenarios; In terms of accuracy experiments, if multiple users can see the anchor point, the average error is 2.7mm, if the anchor point is partially occluded, the average error is 7.16mm, and it can also be observed that the farther the user is away, the average error will become slightly larger, but because they are located within the identification distance of the anchor point, the error is not far away. In different application scenarios, there will be results that are not far away, so the launch of this platform system is very feasible for HoloLens 1, and can be appropriately modified and ported to other application scenarios.