The effect of image variations on fractal dimension calculations

Abstract

Objectives. We used digitized dental radiographs of alveolar bone to test the hypothesis that the fractal dimension, as calculated with the program “ImageFractal” was independent of variations in x-ray exposure, beam alignment, and region of interest placement. Study design. The radiographic data set consisted of 72 radiographs digitized with 200 μm pixels. Radiographs were obtained with the use of three time settings and two alignments. Rectangular regions of interest were placed on each digital image over the interdental bone between the mandibular first and second molars on six hemimandibles. Each of six hemimandibles had identical copies of a unique region of interest placed on every image in its series. New regions were made 3 months later. A fractal dimension was computed from each region of interest with the caliper method included in ImageFractal, a public domain program available through National Institutes of Health. The resulting fractal dimensions were evaluated with two repeated measures analysis of variance. Results. No significant differences were found between the fractal dimensions calculated for baseline images and those from overexposed and underexposed images, from images with 4 to 6 degrees of alignment variations, or from repeat regions of interest. Conclusion. The results support the hypothesis that fractal dimensions derived from digitized dental radiographs are not affected by variations in exposure or small variations in alignment and imply an absolute region of interest placement may not be necessary. However, caution should be used with the use of the fractal dimension to discriminate among alveolar bone variations until further research is performed.