A comparison of cone beam computed tomography and periapical radiography for the detection of vertical root fractures in nonendodontically treated teeth

Abstract

To compare in an ex vivo model, the diagnostic accuracy of periapical radiography and cone beam computed tomography (CBCT) for the detection of artificially induced incomplete and complete vertical root fractures (VRFs), and to determine whether the width of the VRFs had an impact on the diagnostic accuracy of the imaging systems. Incomplete VRFs were induced in 30 nonendodontically treated human mandibular premolar and molar teeth. VRF widths were measured using optical coherence tomography. Complete VRFs were induced in 15 of these teeth. 3D Accuitomo and i-CAT CBCT scans and periapical radiographs were taken prior to and after fracture induction. Receiver operating characteristic (ROC) analysis was carried for each imaging technique. In addition, values for sensitivity, specificity, positive and negative predictive values, inter- and intra-examiner agreement were calculated. In the ROC analysis, both CBCT scanners were significantly more accurate than periapical radiography for the detection of incomplete VRFs (P<0.05). The overall area under the ROC curve (AUC) values for 3D Accuitomo, i-CAT and periapical radiography were 0.687, 0.659 and 0.540, respectively. The sensitivity of 3D Accuitomo, i-CAT and periapical radiography was 27%, 28% and 3% respectively. Interexaminer agreement for the detection of incomplete fractures with periapical radiographs, 3D Accuitomo and i-CAT was 0.020, 0.229 and 0.333, respectively. Both CBCT scanners were significantly more accurate (P<0.01) in detecting VRFs of ≥50μm compared with VRFs of <50μm. 3D Accuitomo was significantly better than i-CAT in detecting VRFs of <50μm (P<0.05). For complete fractures, the AUC values for 3D Accuitomo (0.999) and i-CAT (0.998) were significantly higher (P<0.05) than for periapical radiography (0.724). Under the conditions of this ex vivo study, periapical radiographs and CBCT were unreliable for the detection of simulated incomplete VRFs. The widths of the fractures appeared to have an impact on the diagnostic accuracy of CBCT as the detection of VRFs of ≥50μm was significantly higher than those of <50μm. The detection of complete fractures was significantly higher for all systems than that of incomplete fractures.