Quantifying trabecular bone material anisotropy and orientation using low resolution clinical CT images: A feasibility study

• Published in: Medical engineering & physics Vol. 38; no. 9; pp. 978 - 987
• Main Authors: Nazemi, S. Majid, Cooper, David M.L., Johnston, James D.
• Format: Journal Article
• Language: English
• Published: England Elsevier Ltd 01.09.2016; Elsevier
• Subjects: Anisotropy; Cancellous Bone - diagnostic imaging; Clinical CT images; Engineering, computing & technology; Fabric-elasticity equations; Feasibility Studies; Finite Element Analysis; Humans; Image Processing, Computer-Assisted; Ingénierie, informatique & technologie; Radiology; Signal-To-Noise Ratio; Trabecular bone anisotropic elastic properties; Trabecular bone main orientation; X-Ray Microtomography; Trabecular bone anisotropic elastic properties; Fabric-elasticity equations; Trabecular bone main orientation; Clinical CT images
• ISSN: 1350-4533; 1873-4030; 1873-4030
• DOI: 10.1016/j.medengphy.2016.06.011

•Quantified fabric using grey-level structure tensor in upsized micro-CT images.•Derived anisotropic stiffness entries and main orientation using micro finite element.•Fabric explained 94% of the variance in anisotropic stiffness entries.•Fabric predicted main orientation with 4.8° mean error.•It is possible to estimate anisotropy in clinical CT images. Accounting for spatial variation of trabecular material anisotropy and orientation can improve the accuracy of quantitative computed tomography-based finite element (FE) modeling of bone. The objective of this study was to investigate the feasibility of quantifying trabecular material anisotropy and orientation using clinical computed tomography (CT). Forty four cubic volumes of interest were obtained from micro-CT images of the human radius. Micro-FE modeling was performed on the samples to obtain orthotropic stiffness entries as well as trabecular orientation. Simulated computed tomography images (0.32, 0.37, and 0.5mm isotropic voxel sizes) were created by resampling micro-CT images with added image noise. The gray-level structure tensor was used to derive fabric eigenvalues and eigenvectors in simulated CT images. For ‘best case’ comparison purposes, Mean Intercept Length was used to define fabric from micro-CT images. Regression was used in combination with eigenvalues, imaged density and FE to inversely derive the constants used in Cowin and Zysset–Curnier fabric-elasticity equations, and for comparing image derived fabric-elasticity stiffness entries to those obtained using micro-FE. Image derived eigenvectors (which indicated trabecular orientation) were then compared to orientation derived using micro-FE. When using clinically available voxel sizes, gray-level structure tensor derived fabric combined with Cowin's equations was able to explain 94–97% of the variance in orthotropic stiffness entries while Zysset–Curnier equations explained 82–88% of the variance in stiffness. Image derived orientation deviated by 4.4–10.8° from micro-FE derived orientation. Our results indicate potential to account for spatial variation of trabecular material anisotropy and orientation in subject-specific finite element modeling of bone using clinically available CT.