TL;DR
This study investigates how different CT scan resolutions and finite element mesh sizes affect the accuracy of predicting bovine vertebral mechanical properties using QCT/FEA. The findings suggest that while 1mm slice thickness is computationally efficient and yields comparable results to higher resolutions, smaller element sizes (around 0.41x0.41x0.6mm³ or 0.41x0.41x1mm³) are crucial for improving the accuracy of apparent elastic modulus and yield strength predictions.
What Problem Does This Paper Address?
Quantitative Computed Tomography-based Finite Element Analysis (QCT/FEA) is a powerful tool for assessing bone strength, particularly in predicting vertebral fractures, which are a major complication of osteoporosis. However, the accuracy of these QCT/FEA models can be influenced by several factors, including the resolution of the CT scans used to create the models and the size of the elements used in the finite element mesh. This study aims to systematically explore how variations in these two parameters—specifically, CT scan slice thickness (0.6mm vs. 1mm) and element size—impact the predicted mechanical parameters (like elastic modulus, yield strength, strain, strength, and stiffness) of bovine vertebral bodies. Understanding these effects is critical for optimizing QCT/FEA protocols to achieve accurate and reliable predictions of bone strength, balancing computational cost with clinical relevance.
Methods
The researchers used nine bovine vertebral bodies for their study. These specimens underwent CT scanning at two different slice thicknesses: 0.6 mm (referred to as PA resolution) and 1 mm (referred to as PB resolution).
For the QCT/FEA modeling, two main approaches were taken:
- Whole Vertebral Body Models: The CT data was used to reconstruct 3D models of the entire vertebral bodies. These models were then meshed with 8-node linear brick elements (C3D8). Material properties were assigned based on the relationship between CT greyscale values and bone density, assuming transverse anisotropy. These models were used to predict principal strains and were validated against experimental compressive mechanical tests.
- Cancellous Bone Models (Cuboid VOI): To specifically investigate the effect of element size on trabecular bone, a cuboid region of interest (VOI) was cropped from the center of each vertebral body. These VOIs were then meshed using hexagonal structural elements with six different element sizes, ranging from 0.41x0.41x0.41 mm³ to 3x3x3 mm³. These models were subjected to axial compression simulations to determine apparent elastic modulus and yield strength.
The study then compared the mechanical parameters derived from models created with different scan resolutions and element sizes, using linear regression and Bland-Altman plots for statistical analysis.
Datasets
The study utilized CT scan data from nine bovine vertebral bodies. These were scanned using a clinical CT scanner (Somatom Sensation 64, Siemens) with specific parameters (120kV, 260 mAs, 0.41x0.41 mm/pixel resolution). The datasets were reconstructed at two slice thicknesses: 0.6 mm and 1 mm. No new datasets were introduced; rather, existing CT data was manipulated to explore the effects of resolution.
Key Insights and Ideas
- Scan Resolution vs. Computational Cost: While a higher scan resolution (0.6mm slice thickness) was used, the study found that mechanical parameters predicted from models with a 1mm slice thickness were remarkably similar. This is significant because 1mm slice thickness is more computationally efficient, suggesting it's a viable option for QCT/FEA without a substantial loss in accuracy for whole vertebral body models.
- Element Size Matters for Trabecular Bone: For the cancellous bone models, element size had a more pronounced effect. Smaller element sizes, particularly around 0.41x0.41x0.6 mm³ and 0.41x0.41x1 mm³, led to higher accuracy in predicting apparent elastic modulus and yield strength compared to larger element sizes.
- In-Plane Resolution Dominance: The study hints that in-plane resolution might have a greater impact on mechanical parameter predictions than slice thickness, suggesting that optimizing pixel resolution is crucial.
- Model Validation: The QCT/FEA models were validated against experimental mechanical tests, showing good correlations for principal strains (R² > 0.7) and modest correlations for overall vertebral strength and stiffness (R² > 0.6).
Key Results
- Whole Vertebral Body Models:
- Predicted principal strains from both 0.6mm and 1mm resolution models showed significant linear correlations with experimental measurements (R² > 0.7, P < 0.0001).
- Vertebral strength and stiffness predicted from the two resolutions were similar, with high correlation coefficients (R² = 0.992 for strength, R² = 0.962 for stiffness).
- The computational cost was lower for the 1mm resolution models.
- Cancellous Bone Models:
- Models with element sizes of 0.41x0.41x0.6 mm³ and 0.41x0.41x1 mm³ showed higher accuracy in predicting apparent elastic modulus and yield strength compared to larger element sizes.
- Larger element sizes (e.g., >1x1x1 mm³) led to a significant decrease in apparent elastic modulus and increased computational errors.
- The apparent elastic moduli and yield strengths of models reconstructed from the two scan resolutions with the same element size were generally similar.
Limitations and Future Work
The authors acknowledge several limitations:
- Material Property Differences: The study used a density-elasticity relationship derived from human vertebral bodies for bovine specimens. This mismatch in material properties could introduce inaccuracies. Future work could involve developing specific relationships for bovine bone.
- Limited Range of Scan Resolutions: The study only compared two slice thicknesses (0.6mm and 1mm). A wider range of scan resolutions might reveal more significant differences.
- In Vitro vs. In Situ: The study used ex vivo (in vitro) specimens. In vivo scans might have different characteristics due to factors like patient positioning and tissue attenuation, which could influence FEA results.
Future work could focus on exploring a broader spectrum of scan resolutions and element sizes, investigating the impact of different material property assignments, and extending these findings to in vivo QCT/FEA applications for human bone analysis.
Summary updated