開發交替式動態X光二維/三維對位方法以量化自行車踩踏時下肢軟組織移動誤差及其對膝關節力學分析之影響

Abstract

基於立體攝影術量測人體動作已發現黏貼在皮膚表面的反光球相對其下骨頭的位移，是動作分析中主要的誤差來源之一，稱之為Soft tissue artefacts (STA)。欲利用立體攝影術分析自行車運動亦受此誤差影響無法達到準確的量測，而現今亦缺乏對自行車運動系統性的STA分析結果限制了結果的正確解讀。為了精確的量測關節運動，使用受試者個人化模型結合雙平面動態X光的對位技術量測是目前較好的做法，然而現今並沒有針對臨床常見交替式拍攝的雙平面系統提出適合的對位技術。因此，本研究旨在發展與驗證可應用於雙平面交替式動態X光影像系統之比對技術，並結合立體攝影術的拍攝結果應用在量化自行車運動時的STA，其及對估算骨頭方位、計算膝關節運動學與力動學參數之影響。 本研究提出了三交替影像對位法的比對流程，針對僅雙平面系統與結合輔助量測系統的拍攝條件設計三種運動學模型（等速度運動學模型、剛體運動學模型與類剛體運動學模型）搭配三交替影像對位法用以預測前後影格的骨頭方位，經試體實驗驗證不同速度與運動模式下的對位結果，對比於單平面與雙平面對位，三種運動學模型在動態過程皆有良好的對位結果，所提出基於等速度運動學模型的快速修正演算法已經可以降低單平面對位89%的出平面誤差量，讓三維對位誤差降至小於0.7 mm。經最佳化收斂後三種運動學模型的誤差結果可達到比擬同步雙平面的對位結果，並應用剛體運動學模型量測活體膝關節運動以量化自行車的STA。 自行車運動大腿上的反光球比起小腿有更多的STA，而且更容易受到踩踏阻力的影響。STA會隨著鄰近關節角度的改變而跟著變化，小腿多呈線性大腿則以非線性居多。黏貼在近關節處的反光球比近肢段中段處的反光球有更大的移動範圍，但卻有較小的變化量。估算肢段方位上，STA對大腿及小腿主要造成剛性移動與旋轉，非剛性移動較小。非剛性移動範圍的大小不能代表該反光球組所估算骨頭方位的準確度。力學分析上，阻力並不影響關節角度的誤差範圍，使用兩組反光球組分開計算肢段的方向與位置，有助於同時降低關節角度與力矩的誤差範圍。 本研究建構的方法有助於臨床系統與研究結合，量化的自行車STA結果有助於未來自行車運動的結果解讀，收取的實驗資料將可供後續相關研究使用。

Soft tissue artefacts (STA) have been recognized as a major source of error as applying stereophotogrammetry for human movement analysis. It not only affects the measurement of cycling motion but also limits interpretation of the results from the stereophotogrammetry-based measurement system. Currently, study of STA absent detailed, substantial results which provide guidelines for properly interpret results of cycling. 2D-3D subject-specific model-based registration method combined with biplane fluoroscopy is considered as a non-invasive accurate measurement method. However, the technique design for alternating exposures used in clinical system is not been proposed yet. Therefore, the study aims to develop 2D-3D registration methods for alternating biplane fluoroscopy and used to quantify soft tissue artefacts in the lower limb and their effects on mechanical analysis of the knee during pedalling. A tri-alternating images registration method is proposed combined with three kinematic models (constant speed, rigid and quasi-rigid) which help to predict bone pose of the adjacent frame under sole biplane fluoroscopy or combined with assistant measurement system conditions. The methods were verified by a cadaver study. Compare to single plane and pseudo biplane registration results, fast correction algorithm based on constant speed model already decreased 89% of out-of-plane errors and the target registration error eventually less than 0.7 mm. Performance of three models were comparable to the synchronized biplane registration. The rigid kinematic model was adopted for subsequence in vivo STA quantification during pedalling. Compare to the shank markers, the thigh markers showed greater STA and were affected more by pedal resistance. The STA varied with angles of the adjacent joints, largely linearly for shank markers while non-linearly for thigh markers. Markers near a joint experienced greater ranges of STA than mid-segment markers, but tended to have smaller variation. To estimate bone pose, STA produce greater rigid translations and rotations than the nonrigid component. Range of norigid component may not able to represent accuracy of the marker cluster used to estimate bone pose. On mechanical analysis, calculated joint angles were not affected by different resistant condition. Hybrid two marker clusters to estimate bone orientation and position separately help to reduce error of calculated joint angles and moments. The method developed in the study help to apply clinical sytem for measuring accurate bone kinematics. Results of STA during pedalling help other cycling study and the experiment data will be useful for the further studies.