RESUMO
BACKGROUND: The aim of this study is to determine the influence of field of view (FOV) and number of acquisition projection images (frames) on the detection of chemically simulated peri-implant defects by cone-beam computed tomography (CBCT) using an in vitro bovine rib bone model. METHODS: Eighty implants were placed in bovine ribs in which small and large bone defects were created using 70% perchloric acid. CBCT images were acquired at three acquisition protocols: protocol 1 (FOV 4 × 4 cm, 0.08-mm voxel size, 1,009 frames [high fidelity]; protocol 2 (same as protocol 1 except 512 frames [standard]); and protocol 3 (FOV 14 × 5 cm, 0.25-mm voxel size; high fidelity). Two oral and maxillofacial radiologists (OMRs) and two oral and maxillofacial surgeons (OMSs) rated the presence or absence of bone defects on a five-point scale. κ and area under the curve (AUC) were calculated and compared using analysis of variance with post hoc Tukey test at P ≤ 0.05. RESULTS: Intra- and interobserver agreement for OMRs ranged from moderate to good and from slight to moderate for OMSs. For the detection of small lesions, protocol 1 (AUC 0.813 ± 0.045) provided higher detection rates than protocol 2 (AUC 0.703 ± 0.02) and protocol 3 (AUC 0.773 ± 0.55) [F(2,9) = 1.6377]. For larger defects, the trends were similar, with protocol 1 (AUC 0.852 ± 0.108) providing higher detection rates than protocol 2 (AUC 0.730 ± 0.045) and protocol 3 (AUC 0.783 ± 0.058) [F(2,9) = 1.9576]. CONCLUSION: Within the limits of this study, optimal detection of chemically simulated pericircumferential implant crestal bone defects is achieved at the least radiation detriment using the smallest FOV, the highest number of acquisition frames, and the smallest voxel.