Digital radiography of interproximal bone loss; validity of different filters

The aim of the present study was to compare linear measurements of interproximal bone loss on digitized radiographic images after application of different filters to the gold standard of intrasurgical measurements. Immediately before surgery, 50 radiographs of 50 periodontally-diseased teeth exhibiting interproximal bone loss, were obtained by a standardized technique in 33 patients. Intrasurgically, the distances from the cementoenamel junction (CEJ) to the alveolar crest (AC) and from the CEJ to the deepest extension of the bony defect (BD) were assessed. All radiographs were digitized with a flatbed scanner (resolution: 600x1200 dpi). Using the FRIACOM-soft-ware, the linear distances CEJ to AC and CEJ to BD were measured at 50 interproximal lesions from the digitized but unchanged radiographic images and also after use of 7 different basic image-processing modes (enhancement of contrast [dynamics], inversion, high-pass, enhancement of gray-level differences, mean value, histogram correction, spreading of grey values) with 11-fold magnification. Neither the measurement of the distance CEJ to AC on the unchanged images nor assessments with any of the filters revealed significant differences from the gold standard. Multivariate analysis of variance showed the respective filter (p=0.009), intrasurgical and radiographic assessments (p<0.0001), to statistically significantly, influence the difference between intrasurgical and radiographic measurements of the distance CEJ to BD. The underestimation of the intra-surgically assessed distance CEJ to BD by radiographic measurements ranged from 0.3+/-2.0 to 0.8+/-1.9 mm. The filter "mean value" underestimated interproximal bone loss statistically significantly more than the digitized but unchanged radiograph (p=0.05). In this study, basic digital manipulations (filters) of radiographic images failed to result in statistically significantly more valid measurements of interproximal bone loss when compared to the unchanged but digitized images. All radiographic assessments on the digitized images except for use of one filter (mean value) came close to the intrasurgical gold standard.