Patient-specific microarchitecture of vertebral cancellous bone: a peripheral quantitative computed tomographic and...

This study directly compares peripheral quantitative computed tomography (pQCT) and histology for the assessment of 11 morphological parameters. Sixty-eight cylindrical cancellous bone samples were cored from the thoracic (T-9) thoracolumbar (T-12 or L-1), and lumbar (L-4) vertebral bodies of nine autopsy subjects (aged 44-88 years). Four transverse slices were acquired by pQCT from the bottom to the top of each cylinder. Slice thickness was 300 microm and pixel size was 70 x 70 microm. Thin sections (5 microm) were obtained at the same location in the samples, stained with Von Kossa, and photographed. Classical morphological parameters and strut analysis parameters were measured on all images (272 pQCT and 272 matched histological sections). Because of the partial volume effect and specific thresholding procedure, pQCT overestimated the absolute value of the bone volume fraction (BV/TV) and trabecular thickness (Tb.Th) by a factor 2. The trabecular number (Tb.N), trabecular spacing (Tb.Sp), and total strut length (TSL) were correctly estimated. However, the direct correlation between pQCT and histology was excellent (r2 > 0.85, p < 0.001) for BV/TV, Tb.N, Tb.Sp, TSL, and star surface. For Tb.Th, number of nodes, and number of free ends, the correlation was also good (r(2) > 0.6, p < 0.001). Using a random regression model, we also explored the ability of these parameters to add structural information to the readily available BV/TV or apparent density. The model identified significant (p < 0.001) differences between subjects. For a given BV/TV, some patients had more trabeculae (Tb.N) that were thinner (Tb.Th) and more disconnected (higher free ends and star). This was observed for both histology and pQCT morphometrical data. Our analysis demonstrates the capacity of both histology and pQCT to detect subjects with specific structural patterns in vertebral cancellous bone.