Relationship Between Computed Tomographic Image Analysis and Histomorphometry for Microarchitectural Characterization of Human Calcaneus

The present study aimed to characterize the relationships between several variables reflecting bone microarchitecture assessed by both computed tomographic (CT) image analysis and histomorphometry (conventional CT system) at the calcaneus. A total of 24 cadaveric specimens were studied. The mean age at death was 78 ± 10 years (range, 53–93 years). A total of 15 sagittal sections (1 mm in width and spaced 2 mm apart) were selected for CT analysis; 6 undecalcified sections (7 μm) were analyzed for histomorphometry. The histomorphometric analysis was performed on a Leica Quantimet Q570 image analyzer. Features measured by both methods were: bone volume/tissue volume (BV/TV), trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), trabecular number (Tb.N), interconnectivity index (ICI), number of nodes (N Nd), number of terminus (N Tm), node-to-node strut count (NNS), node-to-terminus strut count (NTS), terminus-to-terminus strut count (TTS), marrow space star volume (SV), Euler number (EN), and fractal dimension (FD). The coefficient of correlations’ values (simple linear regression) between histomorphometry and CT image analysis varied according to the parameters selected. R values were high for BV/TV, Tb.N, and Tb.Sp (range, 0.69–0.90; P < 0.01). R values were less significant for some variables also obtained from the binary image: SV (0.5, P < 0.05) and EN (0.43, P < 0.05). Finally R values were also significant for (two) variables obtained from skeletonized images, i.e., N Nd (0.4, P < 0.05) and N Tm (0.61, P < 0.01). Other correlations were not statistically significant. Moreover, for some variables the relationships between the two methods (CT analysis and histomorphometry) seemed best-described by using nonlinear models. For example, a logarithmic model was more appropriate for SV (r = 0.71, P < 0.01), N Nd (r = 0.52, P < 0.01). Finally the relationship between apparent (App) N Tm and N Tm was most satisfying when using an exponential model (r = 0.64, P < 0.01). In conclusion, trabecular bone structure measures determined on CT images show highly significant correlations with those determined using histomorphometry. The level of correlation varies according to the type of method used for characterizing bone structure, however, and the strongest correlations were found for the most basic features (Parfitt’s parameters). Finally, for some variables, nonlinear models seem more appropriate.