發展人體膝關節統計形狀模型以利三維動態X光量測關節運動

Development of A Statistical Shape Model of the Human Knee for Three-Dimensional Fluoroscopic Imaging of the Joint Motion

李松穎(Song-Ying Li)

指導教授 ： 呂東武

共同指導教授 ： 林正忠

國立臺灣大學/工學院/醫學工程學研究所/碩士(2018年)

https://doi.org/10.6342/NTU201802473

摘要

基於雙平面動態X光影像進行比對得到運動學資訊在目前為活體非侵入性方法中較精確且常見的，過去已有許多文獻應用相關技術於運動學的量測，而在這樣的技術中，最重要的部分就是要先重建個人化三維骨模型，其也能被應用於許多臨床相關之研究，而在目前最精確的方法是透過電腦斷層掃瞄影像進行重建，然而其會有輻射劑量高與花費高等問題。因此，發展替代技術對於骨科相關之研究是必要且有幫助的，在過去文獻中雖有利用二維X光影像搭配如統計形狀模型等變形模型方法建構三維骨頭表面模型，然而其重建誤差普遍都偏高，是否能達到三維動態X光量測之高精度需求是很值得探討的。
本研究以五十七隻膝關節骨模型建立膝關節統計形狀模型，並開發搭配動態X光影像重建個人化膝關節骨模型之方法，透過最佳化的方法進行影像比對以重建骨模型，並透過電腦模擬驗證、試體實驗驗證與活體實驗，以驗證方法之可行性、量化三維骨模型重建誤差以及其應用於膝關節運動學比對之結果。
其結果顯示在活體膝關節骨模型重建誤差上，股骨與脛骨模型分別有0.69 ± 0.09與0.7 ± 0.07 mm之良好的方均根誤差，而透過重建模型所比對之結果相對於CT模型之結果，其所有影像平均比對誤差之平均於股骨為0.60 ± 0.09 mm，於脛骨為0.64 ± 0.11 mm，有高的比對精確度。本研究所開發之方法能以高準確度重建膝關節骨模型，且具有足夠能力應用於動態X光量測關節運動上。

關鍵字

個人化骨模型重建 ； 統計形狀模型 ； 動態X光 ； 影像比對法 ； 膝關節 ； 運動學

並列摘要

Measuring joint kinematics via bi-plane fluoroscopic images is a common and accurate in-vivo non-invasive method. The most important part above is the subject-specific bone model reconstruction. Subject-specific bone models are required for many clinical and biomechanical applications. Currently, the most accurate bone models can be obtained from computed tomography (CT). However, it may cause some concerns such as radiation exposure. Therefore, developing an alternative method is necessary. The previous studies showed higher reconstruction error, and whether it can achieve the high precision requirement of 3D fluoroscopy measurement is very worthy of discussion.
This study collected 57 models of knee joint to establish the knee joint statistical shape model (SSM), and develop a method for reconstruction of subject-specific knee joints via the fluoroscopic images. The three-stage experiments inclusive of computer simulation, in-vitro and in-vivo experiments are performed to evaluate the constructing errors and access the performance in 3D kinematics measurements.
The results showed 0.69 ± 0.09 mm reconstruction error on femur models and 0.7 ± 0.07 mm on tibia models for in-vivo knee joints reconstruction. The average of the mean target registration errors of all frames relative to CT models are 0.60 ± 0.09 mm on femur models and 0.64 ± 0.11 mm on tibia models. This indicated that the method developed by this study can reached the high accuracy for knee joint bone reconstruction, and can be applied on the joint kinematics measurement via fluoroscopy.

並列關鍵字

subject-specific bone reconstruction ； statistical shape model ； fluoroscopy ； image registration ； knee joint ； kinematics

參考文獻

1. Gamage, P., et al. 3D reconstruction of patient specific bone models from 2D radiographs for image guided orthopedic surgery. in Digital Image Computing: Techniques and Applications, 2009. DICTA'09. 2009. IEEE.

2. Krekel, P.R., et al. Interactive simulation and comparative visualisation of the bone-determined range of motion of the human shoulder. in SimVis. 2006.

3. Siston, R.A., et al., Surgical navigation for total knee arthroplasty: a perspective. Journal of biomechanics, 2007. 40(4): p. 728-735.

4. Fernandez, J., et al., Anatomically based geometric modelling of the musculo-skeletal system and other organs. Biomechanics and modeling in mechanobiology, 2004. 2(3): p. 139-155.

5. Harrysson, O.L., Y.A. Hosni, and J.F. Nayfeh, Custom-designed orthopedic implants evaluated using finite element analysis of patient-specific computed tomography data: femoral-component case study. BMC Musculoskeletal Disorders, 2007. 8(1): p. 91.