Texture analysis of high-resolution computed tomograms as a supplementary feature in the diagnosis of osteoporosis: in vitro studies on vertebral specimens.

Purpose: The purpose of this study was to perform texture analysis of high- resolution CT images obtained from human vertebral specimens and to correla te these parameters with the biomechanical stability of the specimens. In a ddition, structure data were compared with bone mineral density (BMD) asses sed by quantitative CT (PCT). Material and Methods: High-resolution CT imag es and standard PCT sections were obtained in 36 vertebral motion segments, each consisting of two vertebrae with intact ligaments and intervertebral disc. The trabecular structure in the CT images was assessed using three te xture analysis techniques: Trabecular threshold area ratio (TTAR), fractal dimension without thresholding (OTS) and fractal dimension with thresholdin g (ITS). Finally, the maximum compressive strength (MCS) was determined usi ng a biomechanical testing device. Results: A correlation of r = 0.76 (p < 0.01) was obtained for TTAR versus MCS, of r = 0.83 (p < 0.01) for ITS vers us NICS, and of r = 0.35 (p > 0.01) for OTS versus MCS, while r = 0.76 (p < 0.01) was found for BMD versus MCS. Best results were obtained by combinin g structure measures and BMD (r = 0.85, p < 0.01). Conclusions: This in vit ro study showed a significant correlation between structure measures and bi omechanical strength, which was comparable to BMD and strength. However, be st correlations were obtained by combining both measures. Using both BMD an d structure measures therefore may improve the prediction of biomechanicall y determined bone strength.