Introduction: Patient-specific evaluation of bone mechanical properties usually relies on the apparent density using clinical Quantitative Computed Tomography (QCT). However, evidence has been reported that bone microarchitecture has an important effect on its mechanical properties.
Objective: To study the feasibility of predicting bone morphology from QCT using multi-scale image analysis.
Materials and Methods: Two micro computed tomography (μCT) images of a porcine vertebral body were computed. The resolution of the μCT images was reduced to 0.3 mm using a corrected average algorithm in order to get the corresponding macro CT images. A multi-scale geostatisitical simulation algorithm was used to generate 100 equally probable μCT scenarios using only the CT image. Two morphometric parameters were then measured on the original and the simulated μCT images: bone volume (BV) and bone surface (BS). The estimation error of these parameters was then computed.
Results: Morphology of trabecular bone could be predicted with an average error of 3% (0.5% to 9.5%). The evaluation of each parameter separately demonstrated that BS had the most error. It could be predicted with an average error of 3.1% (0.5% to 7.0%) while BV could be predicted with an average error of 0.4% (1.6x10-14% to 1.2%).
Conclusion: The results show that the proposed method allows predicting bone morphology. However, improvements of the algorithm are required to automatically identify which predicted μCT image is the most similar to the target one.
Significance: The non invasive estimation of trabecular architecture using QCT may prove to be useful for the prevention and treatment of bone pathologies.