Cone-Beam Images Detect Proximal and Occlusal Surface Caries

Objectives: Conventional radiographic imaging of teeth underestimates the presence of caries. The recent availability of high-resolution cone-beam imaging offers the prospect for improved caries detection. The objective of this study was to compare the accuracy of high-resolution cone-beam images with conventional CCD images for detecting proximal and occlusal dentinal caries. Methods: Non-restored, extracted human permanent premolar and molar teeth were mounted and then imaged with a 3DX Accuitomo (Morita) and a CCD sensor (E2V Technologies). We selected 92 occlusal and 100 proximal surfaces for observers to score. Of these, 36 and 25 respectively, had lesions extending into dentin. Gross lesions were excluded. Using a 5-step confidence scale, five practicing dentists evaluated the images for the presence of caries in dentin using both modalities. Actual presence and extent of caries was established with microCT images. The primary data were collapsed to a binary judgment for the presence or absence of caries. McNemar's test was used to detect differences between imaging modalities for each observer and a paired t-test was used to detect differences between mean values for each modality (p ≤ 0.05). Results: For detecting proximal dentinal caries, four of our five observers had significantly higher sensitivity values with 3DX than with the CCD images. All observers had comparable specificity values. The accuracy with 3DX (0.87) was significantly higher than with the CCD images (0.80). For detecting occlusal dentinal caries, four of five observers had significantly higher sensitivity values when using the 3DX. However, four of five also had significantly lower specificity scores with the 3DX. The accuracy with 3DX (0.79) was not significantly different than with the CCD (0.82). Conclusions: Practicing dentists were able to improve their accuracy in interpreting proximal dentinal caries, but not occlusal dentinal caries, with 3DX high-resolution cone-beam images compared to CCD images.