Utility and effectiveness of computer-aided diagnosis of dental caries.

ABSTRACT

Digital radiography has created a growing opportunity for computer-aided diagnostic (CAD) tools. The Logicon Caries Detector (LCD), with upgraded CAD software based on user feedback, was re-evaluated for its effectiveness via a retrospective clinical study. Using the upgraded LCD software, 12 dentists (evaluators) blindly assessed 17 radiographs taken by another (attending) dentist, who restored 28 proximal surfaces. The attending dentist confirmed the presence of early dentinal caries, as well as identifying 48 surfaces as caries-free or with enamel caries only subject to noninvasive treatment. The radiographs, imported into the software using a digital imaging and communications in medicine (DICOM) reader, were visually assessed under typical operatory lighting conditions, then with the aid of the software's density analysis tool. The effectiveness of the evaluators was gauged by calculating two measures of performance, sensitivity and specificity, for the detection and classification of dentinal caries. Sensitivity among all evaluator dentists was 30% with the initial image; 34% with the brightness and contrast adjusted image; 39% when the image was sharpened; and 69% when the density analysis tool was utilized. Specificity was found to be 97% with the initial image; 95% with the brightness and contrast adjusted image; 93% with the sharpened image; and 94% when the density analysis tool was used. Compared to the unaided eye, the LCD can significantly improve dentists' ability to detect and classify caries. Dentists may be able to find twice as much early dentinal caries requiring restoration (or at least aggressive noninvasive treatment) than previously, while not unnecessarily restoring additional healthy teeth. The LCD enables dentists to obtain more information from dental digital radiography than is possible with the unaided eye, leading to improved patient care.