Characterizing Trabecular Bone structure for Assessing Vertebral Fracture Risk on Volumetric Quantitative Computed...

Area of Science:

• Biomedical Engineering
• Radiology
• Orthopedics

Background:

• Proximal femur bone mineral density (BMD) is standard for fracture risk, but spinal BMD shows stronger links to spinal fractures.
• Volumetric quantitative computed tomography (vQCT) highlights the importance of trabecular bone structure.

Purpose of the Study:

• To assess spinal vertebrae fracture risk using advanced trabecular bone structure features.
• To evaluate the efficacy of second-order statistical features from CT images for predicting vertebral failure load.

Main Methods:

• Axial multi-detector CT (MDCT) images of 28 spinal vertebrae specimens were analyzed.
• Trabecular bone regions were isolated, and BMD and six Gray-Level Co-occurrence Matrix (GLCM) texture features were extracted.
• A generalized radial basis functions (GRBF) neural network predicted specimen failure load, with performance measured by root-mean-square error (RMSE).

Main Results:

• The GLCM texture feature 'correlation' achieved the best prediction performance (RMSE = 1.02 ± 0.18).
• 'Correlation' significantly outperformed other GLCM features (p < 0.01) and mean BMD (RMSE = 1.11 ± 0.17, p < 10⁻⁴).

Conclusions:

• Characterizing trabecular bone structure with GLCM texture features significantly enhances biomechanical strength prediction in spinal vertebrae.
• This approach offers a more accurate method for spinal fracture risk assessment compared to traditional BMD measurements.