Imaging of mandibular trauma: ROC analysis

Objectives: To compare the diagnostic efficacies of panoramic radiographs, mandibular trauma series, and digitized radiographs for detection of fractu res of the mandible. Methods: Fractures were induced using blunt trauma in 25 cadaver mandibles. Panoramic radiographs, digitized radiographs, and man dibular series comprising an anteroposterior view, two lateral oblique, and a reverse Towne's were used. Six observers recorded their diagnoses using a five-point confidence rating scale. The data were analyzed using receiver operating characteristic (ROC) curve analysis. Results: Significant differ ences based on imaging modalities were found (p < 0.0015) in the area under the curves (A(z)): panoramic radiograph, 0.8762; mandibular series, 0.7521 ; panoramic plus anteroposterior radiographs combination, 0.8886; and digit ized mandibular series, 0.7723. No observer-based differences were noted. C ondylar and coronoid fractures were difficult to detect than those in other areas of the mandible (p < 0.033). Intra- and interobserver agreements wer e high (kappa (W) = 0.81 and 0.76, respectively). Conclusions: Panoramic ra diographs are adequate for detection of mandibular fractures. Addition of a n anteroposterior view augments diagnostic accuracy.