Evaluating Artificial Intelligence in Full-Arch CBCT Caries Detection: A Comparative Analysis with Clinical Assessment.

Objectives: The aim of this study was to compare the diagnostic accuracy of the Diagnocat system (DGNCT LLC, Miami, Florida, USA), based on artificial intelligence algorithms, with clinical assessments performed by three dentists. Materials and Methods: The analysis was based on data obtained from cone-beam computed tomography (CBCT), focusing on the detection of carious lesions. The inclusion of three specialists with comparable levels of knowledge and professional experience increased the reliability of the results. The dentists classified teeth with carious lesions solely on the basis of CBCT imaging, physical examination, and their own clinical knowledge, under single-blind conditions, without awareness of the subsequent comparative analysis. Results: The results demonstrated a variable level of agreement between the Diagnocat system and the dentists' assessments, depending on factors such as tooth location, as well as patient age and gender. The lowest level of agreement was observed in premolars, which may be attributed to their complex morphology. Higher diagnostic accuracy was noted in molars and incisors, particularly in younger patients. Conclusions: Further research should focus on the integration of various diagnostic modalities, including diagnostic imaging, intraoral scans, and photographic documentation, which may significantly enhance diagnostic precision, especially in cases of early-stage lesions. According to the results, the Diagnocat system demonstrates potential as a supportive tool in the diagnostic process in dental practice, as well as a screening tool enabling preliminary evaluation of imaging studies.