Fractal dimension calculated from two types of region of interest

Abstract

Objectives: To compare the effect of using two region of interest (ROI) types on calculating the fractal dimensions of trabecular bone during simulated osteoporosis. Methods: Ten 5 mm thick cross-sections from the long bone of a cow were progressively decalcified in 0.1 N Hcl for 5, 10, 20, 30, 60 and 90 min intervals, and radiographed using 0 °projection angle in a specially designed device. Two types of ROI (the ruggedness of the boundary and bone profiles) were placed on each digital image. Fractal dimensions and variance in mean pixel intensity were computed from each ROI using the caliper method in ImageFractal (National Institutes of Health, Washington, DC, USA). Correlation analysis quantified the relationship between changes in variance and fractal dimensions of the two types of ROI. Results: A Strong correlation (r = 0.90 ~ 0.98, P ≤ 0.0026) was found between variance in mean pixel intensity and decreasing fractal dimensions calculated from bone profiles. There was no consistency in the ruggedness of the boundary and increasing decalcification. Conclusions: The fractal dimensions used to characterize the ruggedness of a boundary of trabecular bone in a two-dimensional ROI are inadequate for the detection of osteoporosis, but those calculated from bone profiles may be a sensitive descriptor of trabecular bone structure.