Correlation between diagnostic accuracy and perceptibility

Abstract

Objectives: To correlate diagnostic accuracy for proximal caries with perceptibility of low contrast image details using regression analysis. The other purpose was to determine the attenuation range required for proximal caries diagnosis. Methods: The results of the two types of observer performance tests described above were retrieved from previous studies. Recording media included in those studies were the Compuray and the Dixel, direct digital radiographic systems, and Ektaspeed Plus film. The average numbers of perceptibility of image details from five observers were calculated for each step and for every combination of contiguous steps of the aluminium test phantom from the perceptibility test. The average diagnostic accuracy for proximal caries from the same five observers was correlated with the total number of perceptible details from the phantom using regression analysis. Finally, attenuation range required for proximal caries diagnosis was calculated from the attenuation range of the phantom where the maximum correlation coefficient was obtained. Results: Maximum correlation (r = 0.68) was obtained at the combination of five contiguous steps of the aluminium test phantom. Attenuation range required for proximal caries diagnosis corresponded to the 2 mm to 6 mm thickness of aluminium with acrylic block of 12 mm thickness. Conclusions: There is a correlation between perceptibility of low contrast image details and diagnostic accuracy for proximal caries. There may be a possibility to simplify observer performance tests for proximal caries diagnosis by using the standardized phantom simulating its attenuation range. © 2005 The British Institute of Radiology.