Accuracy of caries diagnosis in digital images from charge-coupled device and storage phosphor systems: an in vitro study.

Publisher Website
Google Scholar
Abstract
To evaluate the diagnostic accuracy of carries detection with four intra-oral digital radiographic systems in vitro and to investigate the impact of image compression. 116 extracted human premolars and molars mounted three in a line with approximal contacts were radiographed with four digital systems (Digora, RVG, Sens-A-Ray, and Visualix) and imported into a Windows-based program. Sixteen images from each system were compressed (JPEG, irreversible compression). The total of 528 images was assessed by six radiologists using a 5-point confidence scale for the detection of approximal and occlusal caries. Caries was validated histologically. The disease threshold was caries in enamel for approximal surfaces and in dentine for occlusal surfaces. Mean ROC curve areas for approximal surfaces were 0.611 (DIG), 0.572 (RVG), 0.594 (SAR), and 0.596 (VIX), and for occlusal surfaces 0.794 (DIG), 0.819 (RVG), 0.751 (SAR), and 0.761 (VIX). There were no significant differences between the areas under the ROC curves (p > 0.05); nor were there any significant differences between the compressed and uncompressed images (p > 0.05). The four digital systems performed almost equally well for detection of caries in vitro, and compressed images were as accurate as uncompressed.