Longitudinal analysis of radiographic trabecular pattern by image processing.

To describe structural and textural changes in bone structure, the radiographic trabecular pattern of children, aged 4-14 years, was examined using a digital image processing system. This investigation is based on data from the Nijmegen Growth Study, a mixed-longitudinal growth study comprised of three birth cohorts, which were observed for five years. Of 3075 left hand-forearm radiographs of 426 children, a standard area of 10 x 10 mm of the radius was digitized and the trabecular pattern was described by ten image features. It is demonstrated that all image features show significant changes during the observation period (multivariate analysis of variance of p < 0.01). The age of the children correlates significantly with the image features (Spearman's Rho = 0.4; p < 0.05). Because of cohort effects, the three cohorts were studied separately. Highest correlations between age and trabecular pattern were shown in the cohort of the youngest children. The changes of the image features during the observation period seems to indicate that the fine trabecular pattern of young children changes into a coarser adult pattern. It is concluded that digital quantification of the radiographic trabecular pattern is appropriate to describe changes in trabecular bone.