OBJECTIVE: The purposes of this study were to examine the density correction afforded by curve-fitting algorithms and to investigate whether the device we developed significantly improves the reliability of longitudinal alveolar process bone radiographic density measurements. STUDY DESIGN: Stepwedges were radiographed over a range of impulse settings, and curve-fitting algorithms were fitted to sets of step images on each digitized film. Differences between the actual thicknesses of an alternate set of steps and their corresponding thickness estimates were calculated. Next, clinicians made periapical radiographs from interproximal bony sites on a dry skull using our imaging device. Differences in bone densities between corresponding regions of interest taken 1 week apart were calculated. RESULTS: Analysis of variance and Duncan's Multiple Range test demonstrated that piecewise linear, third-degree polynomial, and fourth-degree polynomial curves provided significantly better estimates of stepwedge thickness than did sigmoid or first degree polynomial-curves (P < .05) and that the differences between repeat bone density measurements made with density correction were significantly less than those made without density correction (P < .05). CONCLUSIONS: Piecewise linear, third-degree polynomial, and fourth-degree polynomial curve-fitting algorithms provided the best densitometric correction. The use of our imaging device increased the reliability of longitudinal bone density measurements.
OBJECTIVE: The purposes of this study were to examine the density correction afforded by curve-fitting algorithms and to investigate whether the device we developed significantly improves the reliability of longitudinal alveolar process bone radiographic density measurements. STUDY DESIGN: Stepwedges were radiographed over a range of impulse settings, and curve-fitting algorithms were fitted to sets of step images on each digitized film. Differences between the actual thicknesses of an alternate set of steps and their corresponding thickness estimates were calculated. Next, clinicians made periapical radiographs from interproximal bony sites on a dry skull using our imaging device. Differences in bone densities between corresponding regions of interest taken 1 week apart were calculated. RESULTS: Analysis of variance and Duncan's Multiple Range test demonstrated that piecewise linear, third-degree polynomial, and fourth-degree polynomial curves provided significantly better estimates of stepwedge thickness than did sigmoid or first degree polynomial-curves (P < .05) and that the differences between repeat bone density measurements made with density correction were significantly less than those made without density correction (P < .05). CONCLUSIONS: Piecewise linear, third-degree polynomial, and fourth-degree polynomial curve-fitting algorithms provided the best densitometric correction. The use of our imaging device increased the reliability of longitudinal bone density measurements.