Does anthropomorphic model design in ex vivo studies affect diagnostic accuracy for dental root fracture using CBCT?

The evidence for diagnostic accuracy using cone beam computed tomography (CBCT) for dental applications depends heavily on ex vivo research, but there is little knowledge of whether the model used affects the diagnostic accuracy results. The objective of this study was to determine the impact of different designs of anthropomorphic models on diagnostic accuracy for the specific task of dental root fracture detection.

Horizontal or oblique root fracture was induced in 24 of 48 permanent maxillary incisors. The 48 teeth were scanned by CBCT using standard clinical exposure factors on five occasions, each with a different model design. Scans were viewed by five dental and maxillofacial radiologists, who each made a forced diagnosis of fracture or no fracture in each root and a judgment on root fracture using a five-point confidence scale. Sensitivity (Se), specificity (Sp) and areas under receiver operating characteristic (ROC) curve (Az) were calculated for each observer for each model.

There were no significant differences between the diagnostic accuracy measurements recorded using different models. There were, however, numerous significant differences between observers using the same anthropomorphic model.

Despite the differences in X-ray attenuation between the five model designs, the results suggest that the anthropomorphic model does not affect the results of diagnostic accuracy studies on root fracture using this CBCT machine at standard clinical exposures. This provides some confidence in the previously published evidence. The interobserver diagnosis differences indicate that research using only two observers could provide misleading results.