Accuracy of Radiographic Shape Assessment of Approximal Caries Lesions

Objectives: To evaluate the accuracy of radiographic depth and shape assessment of caries lesions in approximal surfaces. Methods: Three observers examined jointly radiographs of about 300 human, extracted teeth with approximal caries lesions of known depths (validated by ground section histology) to describe common shapes of enamel and dentine lesions. Four radiographic shape categories for enamel (E1=cup-shaped, E2=triangular, E3=band-shaped, E4=others) and dentine (D1=double cup-shaped, D2=double triangular, D3=band/cup-shaped, D4=others) lesions were described. The inter-observer reproducibility for those categories was tested in radiographs of another tooth material and showed to be substantial. For the final study 93 human, extracted teeth were selected and radiographed under standardised conditions with E-speed films. Three independent observers examined the radiographs for the absence or presence of caries lesions in 177 approximal, unrestored surfaces. If a lesion was detected its depth and shape were assessed. At the end the teeth were serially sectioned and examined histologically for the true presence and depth of caries. For evaluation of the radiographic observer reproducibility 25% of the surfaces were re-examined. Results: Histological caries lesions were present in 131 surfaces (76 in enamel and 55 in dentine). The observers' radiographic lesion depth scores were in accordance with the histological validation in 50-56% of the cases. In 22-32% of the cases the true depth was overestimated and in 14-25% it was underestimated. Histological enamel lesions were predicted most correctly when the shape was E2 and histological dentine lesions when the shape was D3. However, no significant difference was found between the various shape categories, neither for enamel nor for dentine lesions. The inter- and intra-observer reproducibilities were in general low. Conclusion: None of the radiographic lesion shapes tested were highly effective for accurate identification of histologically existing approximal caries lesions, and the observer reproducibility of the tested radiographic shapes was poor.