混合型姿態估測於針具穿刺手術導引的擴增實境系統

Abstract

針具穿刺手術是目前醫療界中相當常見的手術方式，例如射頻燒灼術(RFA) 與活體組織檢查(biopsy)。然而，要讓針具精準地穿刺到指定細胞位置並遵循特定軌跡，醫師們必須在手術過程中採用實時醫療影像或相機影像作為導引，例如超音波影像與針具軌跡預測。然而在針具穿刺過程中，針具軌跡無法被精準預測，這可能會導致手術過程不順暢，甚至給病人帶來危害。其中一大阻礙因素就是針具只能在有限的相機視野內追蹤。因此，我們提出了一種混合性的標記與針具姿態估測擴增實境系統，透過同時追蹤整個針具組合，提升手術的可靠性，並擴展可操作的工作空間。這個系統不僅能夠透過標記追蹤計算手把姿態，還能透過針體的特徵點來估算空間資訊，擴大可偵測的範圍。我們所提出的系統實現了3.13 倍的空間偵測範圍擴展，整體準確度為2.9±0.01mm，此外系統還能夠提供針具彎折的警示，避免器材偏誤造成的手術風險。我們提出的方法使用CCD相機來實現基於AR的手術導引，可以為目前的RFA手術帶來增益效果。

Surgical needle insertion is a commonly performed medical procedure, such as radiofrequency ablation (RFA) and needle biopsy. To ensure that the needle is accurately inserted into the targeted tissue following a designated trajectory, clinicians often rely on real-time image feedback for needle trajectory visualization and prediction using ultrasonography and/or optical camera feedback. However, during needle puncture guiding procedures, the needle trajectory may not always be accurately reported, which can cause challenges or even hazards to the system. One of the major obstacles is that the needle can only be tracked within a limited camera field of view. In this work, we proposed a hybrid pose estimation augmented reality (AR) algorithm that allows for simultaneous tracking of the entire needle set, thus expanding the detectable range of the needle insertion procedure. The algorithm not only tracks the 3D pose of the needle handle (using marker tracking) but also tracks the pose of the needle itself (using key point detection), thereby increasing the trackable zone during the procedure. The detectable view is measured to have up to 3.13 times improvement in the field-of-view, with an overall accuracy of 2.9±0.01 mm. This proposed algorithm represents an advancement in using CCD camera as a tool for implementing AR-based surgical guidance and may be beneficial for current surgical procedures such as RFA or biopsy.