Periodontal disease morbidity quantification. II. Validation of alveolar bone loss measurements and vertical defect diagnosis from digital bite-wing images.

Quantitative alveolar bone loss evaluation by digital processing of dental radiographs was performed. More than 20 contrast and edge enhancement techniques were applied to bite-wing radiographs of adult molar teeth and the results subjectively evaluated. Wallis space variant image enhancement was selected as the most promising candidate for improved visualization of the cemento-enamel junctions and alveolar crests. Seventy-three skulls were radiographed, the films digitized, and enhancement performed on the digital images. Alveolar bone loss measurements from skulls (truth) were compared with those from the original radiographs, unprocessed digital images, and enhanced digital images. The average intraclass correlation coefficient for two observers was 0.83 for radiographs, 0.86 for digital images, and 0.91 for enhanced images. The intraclass correlation coefficient between the 2 sets of dry skull measurements made by two additional observers was 0.88. Means and 95% confidence limits were determined for differences between measurements from dry skulls and radiograph-based images. In the worst case for measurements from enhanced images, the mean difference from truth was -0.40 mm with a 95% confidence interval from -0.54 to -0.25 mm. Receiver operating characteristic (ROC) analysis was used to determine the utility of radiograph-based images for the diagnosis of vertical defects. The ability of two observers to diagnose vertical defects on dry skulls was also tested. All methods of vertical defect diagnosis, including the identification of vertical defects on dry skulls, had relatively low diagnostic performance. Image processing technology has considerable potential in periodontal disease studies for quantitative evaluations of alveolar bone height. The diagnosis of vertical defects using current definitions and techniques is problematic, with considerable room for improvement.