Validation of Material Algorithms for Femur Remodelling Using Medical Image Data

Shitong Luo, Xingquan Shen, Xin Bai, Jing Bai, Jianning Han, Yu ShangApplied Bionics and Biomechanics 2017 · DOI

TL;DR

This paper validates two "error-driven" computational models for bone remodeling using real human femur CT scans. By quantitatively comparing simulated bone density and mechanical properties against medical imaging data, the study demonstrates the potential of these algorithms for predicting bone changes in abnormal mechanical environments and for optimizing implant designs.

What Problem Does This Paper Address?

Bone remodeling, the continuous process of bone resorption and formation, is heavily influenced by mechanical loads. Understanding and accurately simulating this process is crucial for designing better orthopedic implants (like hip or knee replacements) and for developing strategies to combat bone loss, such as that experienced by astronauts in space due to microgravity. While various computational models (algorithms) exist to describe bone remodeling, many lack rigorous quantitative validation against real biological data. Existing validation methods are often qualitative, making it difficult to assess the true accuracy and reliability of these models. This paper aims to bridge this gap by quantitatively validating two "error-driven" bone remodeling algorithms using medical CT data.

Methods

The researchers employed a combination of the finite element (FE) method and two "error-driven" material algorithms: an isotropic algorithm (treating bone as uniform in all directions) and an orthotropic algorithm (accounting for bone's directional properties).

  1. FE Model Construction: A 3D FE model of a human femur was reconstructed from CT scan data. This model was meshed into eight-node hexahedron elements, with element properties (like density) derived from the CT Hounsfield Units (HU).
  2. Material Algorithms:
    • Isotropic Algorithm: This algorithm assumes bone is uniform in all directions. The mechanical stimulus (a strain-based variable) drives changes in Young's modulus.
  3. Orthotropic Algorithm: This algorithm is more complex, accounting for bone's directional properties. It involves two key steps:
  4. Determination of Orthotropic Axes: This step was an improvement on a previous 2D model, now considering multiple load cases and daily load cycles to determine the principal material directions for each element.
  5. Bone Stiffness Modification: Similar to the isotropic algorithm, changes in Young's moduli along the principal axes are driven by the mechanical stimulus and an "error" compared to a reference value.
  6. Simulation Process: The FE model was subjected to simulated hip joint and abductor muscle forces. Iterative calculations were performed to simulate bone remodeling until a steady state (convergence) was reached.
  7. Quantitative Validation: The core of the study involved comparing the simulated bone density and mechanical properties with real data derived from CT scans. Key metrics included:
  8. Correlation Coefficient: Measuring how well the simulated density distribution matched the real distribution.
  9. Mean Error: Quantifying the average difference between simulated and real bone densities.
  10. Local Region Analysis: Comparing densities in specific anatomical regions of the femur.
  11. L-T Ratio (for Orthotropic Algorithm): Evaluating the ratio of longitudinal to transverse Young's modulus, a measure of anisotropy.
  12. Trabecular Alignment (for Orthotropic Algorithm): Visualizing the orientation of the principal material axes, which should align with trabecular bone structure.

Datasets

The study utilized CT data from the Visible Human Project (VHP), specifically a male CT dataset from the National Library of Medicine (NLM), USA. This dataset is described as a standard for human anatomical models and has been used in various biomechanics studies. The researchers also referred to reported L-T ratio data from literature [34] for validation of the orthotropic algorithm's mechanical properties.

Key Insights and Ideas

  • Quantitative Validation is Crucial: The paper strongly advocates for quantitative validation of bone remodeling models, moving beyond qualitative comparisons to ensure their predictive power.
  • "Error-Driven" Algorithms Show Promise: Both isotropic and orthotropic "error-driven" algorithms, when properly parameterized, can produce realistic bone density distributions that align with real CT data.
  • Orthotropic Model Captures Anisotropy: The orthotropic algorithm, by considering material orientation, provides a more nuanced representation of bone structure and mechanical properties, as evidenced by its ability to predict L-T ratios and trabecular alignment.
  • Reference Stimulus is Key: The choice of the "reference stimulus" (Yo) significantly impacts the simulation results. The study identified an optimal Yo value for achieving the best fit with real data.
  • CT Data as a Gold Standard: Medical CT scans, with their ability to quantify bone density, serve as an excellent source for validating computational bone remodeling models.

Key Results

  • Both the isotropic and orthotropic algorithms produced realistic bone density distributions that generally matched the patterns observed in real CT data.
  • An optimal reference stimulus (Yo = 0.012) was identified, leading to high correlation coefficients (up to 0.76 for isotropic and 0.70 for orthotropic) and relatively small mean errors (0.29 g/cm³ for isotropic and 0.32 g/cm³ for orthotropic) between simulated and real bone densities.
  • The orthotropic algorithm's simulated L-T ratios were close to reported literature values, and the predicted trabecular alignment generally matched expected stress trajectories.
  • The study noted that while simulated bone density distributions were realistic, the transition from high to low density was smoother in simulations compared to the sharper transitions seen in real CT data.

Limitations and Future Work

  • Biological Mechanism Simplification: The "error-driven" algorithm is an "apparent" method that focuses on the relationship between mechanical stimulus and bone amount/structure, but it does not deeply reflect the underlying biological mechanisms of bone remodeling.
  • Non-Site-Specific Reference Stimulus: The current model assumes a uniform reference stimulus across all elements. The authors suggest that a site-specific reference stimulus could improve accuracy, as bone density and mechanical requirements can vary locally.
  • Linear Elastic Material Assumption: The FE analysis assumed bone as a linear elastic material, and marrow was excluded. This simplification might affect results.
  • CT Resolution: While sufficient for geometry and density, the CT spatial resolution (3.0 mm) was not high enough to describe bone microstructure, necessitating the use of literature data for L-T ratio validation.
  • Future Work: The authors propose several avenues for future research:
  • Optimizing Parameters: Assigning optimal, potentially site-specific, reference stimulus values to minimize discrepancies between simulated and real data.
  • Nonlinear Remodeling: Investigating nonlinear bone material algorithms, as bone remodeling is believed to be nonlinear.
  • Integrating Biological Mechanisms: Enhancing models by incorporating more detailed biological processes, potentially by integrating CT data with "reaction-diffusion" or "damage-repair" models.
  • Improving Trabecular Alignment: Further refining the orthotropic algorithm to better capture the sharp transitions in bone density.

Summary updated

Abstract

The aim of this study is the utilization of human medical CT images to quantitatively evaluate two sorts of "error-driven" material algorithms, that is, the isotropic and orthotropic algorithms, for bone remodelling. The bone remodelling simulations were implemented by a combination of the finite element (FE) method and the material algorithms, in which the bone material properties and element axes are determined by both loading amplitudes and daily cycles with different weight factor. The simulation results showed that both algorithms produced realistic distribution in bone amount, when compared with the standard from CT data. Moreover, the simulated L-T ratios (the ratio of longitude modulus to transverse modulus) by the orthotropic algorithm were close to the reported results. This study suggests a role for "error-driven" algorithm in bone material prediction in abnormal mechanical environment and holds promise for optimizing implant design as well as developing countermeasures against bone loss due to weightlessness. Furthermore, the quantified methods used in this study can enhance bone remodelling model by optimizing model parameters to gap the discrepancy between the simulation and real data.
BibTeX
@inproceedings{luo2017validation,
  author = {Shitong Luo and Xingquan Shen and Xin Bai and Jing Bai and Jianning Han and Yu Shang},
  title = {{Validation of Material Algorithms for Femur Remodelling Using Medical Image Data}},
  booktitle = {Applied Bionics and Biomechanics},
  year = {2017},
  doi = {10.1155/2017/5932545},
}
Figure