Subject-Specific Spinal Alignment Analysis in the Supine Position

人體脊椎是一個複雜的結構，它保護脊髓以及傳遞頭顱與軀幹的重量至骨盆。脊椎的穩定度來自於各椎體與周遭的軟組織，而這些軟組織在長期椎體的錯誤擺位下容易受損。人的一生中有三分之一的時間花在睡眠上，因此，在仰臥時維持脊椎的適當位置是非常重要的。然而，人體內力無法直接測得，而因為每個人的背部與身材特徵不同，無法正確求得每個人仰臥時的脊椎型態。進而無法利用個人化分析提供有效的脊椎型態評估與床墊挑選。 本研究以既有之CT資料創建全身有限元素模型，並利用點-多面體測試判斷每一個元素的組織，作為一基底樣版模型（Generic Model）。其次，為建立個人化有限元素模型，本研究使用Kinect量測受試者深度影像，以點雲拼接還原全身人體點雲；並使用非剛體註冊演算法CPD確認受試者表面點雲與CT表面模型對應關係，使用CPD之結果之GMM作Kernal Regression計算受試者體內組織元素節點與CT對應關係。以此可得一個人化表面幾何、體內組織的有限元素模型。最後，結合此全身人體模型與床墊模型，利用有限元素法分析人體仰躺時之脊椎型態，並進行驗證。結果顯示，本研究提出之方法計算之體重誤差在3公斤內；而有限元素分析結果可預測不同床墊下頸部和腰部之角度變化趨勢。本研究提出之方法傾向低估脊椎改變形態之角度，本研究亦提出可能原因與未來可能之改善方法修正。 本方法具有三維量測和非侵入式、無輻射之性質，並提供個人化之分析流程，適用於仰臥時不同床墊材料參數對脊椎型態之分析。本方法將有助於脊椎型態診斷及個人挑選床墊或長期臥床患者之床墊評估與設計。The spine is stabilized by the intervertebral discs themselves and the surrounding soft tissues such as the ligaments and muscles, which are prone to fatigue injuries under prolonged malpositioning. The humans spend about one-third of their lives lying on the bed. Thus, it is essential to maintain a good spinal alignment in the supine position. However, direct measurement of the forces in the body in vivo is possible only in exceptional circumstances. Owing to the varieties of back shapes and body figures across individuals, it is difficult to provide a subject-specific analysis to simulate the spine alignment while lying supine on the mattress. The study reconstructed an existing whole body CT scan data into a three-dimensional model with different kinds of tissues. A generic whole body finite element model was built and each element was assigned a tissue by point-in-polyhedron tests. Secondly, to generate a subject-specific FE model, the study utilized Kinect to capture depth images which were stitched into subject’s whole body point cloud. Surface deformation from CT to subject’s model was determined with non-rigid registration （CPD）. Inner tissue deformation was determined with kernel regression following CPD’s assumption. A subject-specific FE model was then created. Together with the mattress model formed in the software, the study has succeeded in analyzing spinal alignment during supine with validation. The results showed that forecasted subjects’ weight difference with that of real weight were all less than 3 kg, which were accounted for less than 10% of their real weight. The FE analysis results revealed that the study was able to forecast the trend of angle difference around neck and lumbar across different mattress. The study, however, tended to underestimate the deformation of spine, while the possible reason and solutions were also proposed in the study. The study provided a three-dimensional customized process to spinal analysis without invasive technique or radiation. It was helpful to gain insight of spinal alignment during supine with different mattress for different people. The study would help to the development of mattress and the diagnosis of spine