Correction of density changes caused by methodological errors in CADIA

Publisher Website
Google Scholar
Abstract
Quantitative evaluation of radiographic changes by computer-assisted densitometric image analysis (CADIA) requires exact knowledge about method errors. In the present study, density change errors were determined from pairs of films with "no change." From this distribution of error values, a series of threshold values for correction of changes due to method errors was selected. The threshold values were then applied to results from analysis of densitometric changes in sites with "known loss" or "no change" of the alveolar bone among films from cynomolgus monkeys. The density errors formed non-normal distributions with no difference in magnitudes between the absolute values expressing density decreases and density increases in the same areas. Calculation of sensitivity, specificity, type I and type 2 errors showed that these variables were clearly influenced by selection of different threshold values for correction of the density change errors. It is therefore recommended that thresold values be determined for each analytical system initially as well as following any equipment or computer program modification. Before selecting the threshold value for a specific radiographic analysis, the desired level of sensitivity and specificity should be evaluated.