Feasibility of micro‑computed tomography to detect and classify proximal caries lesions in vitro

Background: Micro‑computed tomography (micro‑CT) has been widely described as a nondestructive in vitro imaging method although its accuracy for caries detection is still unclear. The aim of this study was to investigate the feasibility of micro‑CT to detect and classify proximal caries lesions in posterior teeth, using different protocols. Materials and Methods: In this in vitro study, crowns of 122 human teeth were scanned with Skyscan 1174 (Bruker, Kontich, Belgium) using the full‑scan mode (360°). Reconstruction of 900 basis images was performed using NRecon Software (Bruker, Kontich, Belgium) for the protocol 1 (full‑scan mode 360˚), and 450 basis images were used to reconstruct the protocol 2 (half‑scan mode 180°). Three observers analyzed the reconstructed images concerning the presence and depth of proximal caries lesions (244 surfaces). To determine the presence/absence and depth of caries lesions, histological examination was conducted as reference standard level of significance McNemar and McNemar‑Bowker tests compared the methods studied and the gold standard (P < 0.05). Results: The intra‑ and inter‑observer agreement for both methods ranged from moderate to excellent. There was no difference between both micro‑CT methods and histology for the presence of lesions (P > 0.05). However, both methods differed with the reference standard for depth (P < 0.05). The disagreement occurred mostly in cases of enamel lesions. The highest diagnostic values were found for 180° rotation. Micro‑CT performed well in detecting caries lesions compared to histology, meanwhile the classification of their depth presented lower values. Scan mode did not influence the detection. Conclusion: Both protocols of micro‑CT tested presented an overall satisfactory performance in detecting proximal caries lesions; however, for the depth classification, the method was not accurate.