Comparison of reliability of manual and computer-intensive methods for radiodensity measures of alveolar bone loss

To compare the reliability of radiodensity measurements made from dental radiographs with manual and a novel computer-intensive methods.

As part of a prospective study of postmenopausal women, a series of seven vertical bitewing radiographs were taken of 36 patients. One of each set of radiographs was repeated. The original and the corresponding duplicate radiographs were used in this study. Radiographs were digitized at 50 microns spatial resolution and 12-bit gray-scale resolution. For the Manual Method, original and duplicate radiographs were manually cropped to improve image homology, histogram matched and mean pixel gray-scale values determined for an alveolar bone ROI within each image. For the computer-intensive method, images were put into registration with ANALYZE software (Mayo Foundation, Rochester, MINN, USA), cropped automatically, histogram matched and color-coded on the basis of the per cent difference. Alveolar bone ROIs adjacent to clinical crowns and root surfaces whose color code indicated less than a 5% change were sampled. Method error (ME) and the coefficient of variation of method error (CVME) were calculated.

With the Manual Method the SD between original and duplicate measures was 95.21 out of 4096 gray scale values; ME = 67.32; CVME = 3.78%. For the computer-intensive method, the corresponding values were 54.74, 38.71, and 2.29%.

The new computer-intensive method resulted in a 40% improvement over the Manual Method in the precision of radiodensity measurements.