The effect of a knowledge-based, image analysis and clinical decision support system on observer performance in the diagnosis of approximal caries from radiographic images.

The aim of this study was to investigate the effect of a knowledge-based image analysis and clinical decision support system (CariesFinder, CF) on diagnostic performance and therapeutic decisions. The study material consisted of radiographic images of 102 approximal surfaces, 35 sound, 67 caries (25 caries and cavitated and 42 caries). Sixteen general practitioners were presented with (1) radiographic film images and (2) digital filmless images with the results of CF. The viewers were asked to respond whether approximal caries was present and whether a restoration was indicated. Responses were analyzed for accuracy, sensitivity, specificity and agreement. Further, the practitioners were ranked according to the accuracy of their restorative decisions and assigned to ten overlapping groups of 7 practitioners. For each group the diagnostic and therapeutic decisions were then examined for unanimity. The parameters of accuracy, sensitivity and specificity were then established for each group based on only unanimous, correct decisions. The diagnostic and therapeutic accuracy of CF alone was equal or superior to the decisions of the practitioners viewing film images alone. For unanimous decisions, CF alone was more accurate than the most accurate group of practitioners and made fewer incorrect decisions to restore non-cavitated surfaces than the practitioners. In general, dental practitioners viewing the results of CF significantly increased their ability to diagnose caries correctly, their overall diagnostic accuracy, and their ability to recommend restorations for cavitated surfaces. There was a decrease in the accuracy of their restorative decisions overall and in the specificity in particular.