Applicazione della tecnologia di “Intelligenza artificiale” nella gestione della diagnostica per immagini in odontoiatria

Abstract Background: The role of artificial intelligence (AI) in dentistry is becoming increasingly important, particularly in diagnosis and treatment planning. This study aimed to evaluate the clinical performance of AI-driven software (Diagnocat, DGNCT LLC, Miami, US) in analyzing orthopantomograms (OPTs) for patients focusing on its sensitivity, specificity, and precision. Materials and Methods: A total of 104 patients undergoing their first dental consultation at a tertiary care center were included. OPT scans were analyzed by Diagnocat, which assigned a percentage likelihood (0–100%) for caries, bone loss and missing teeth. A board-certified dentist, blinded to Diagnocat’s results, performed independent clinical evaluations using standardized diagnostic criteria. Sensitivity, specificity, and receiver operating characteristic (ROC) curves were computed. Agreement between Diagnocat and expert assessments was evaluated using Cohen’s kappa. Results: Diagnocat exhibited high sensitivity (0.969) but low specificity (0.214) for detecting missing teeth. For caries detection, moderate diagnostic performance was observed (κ = 0.29, sensitivity = 0.805, specificity = 0.591), supporting its potential in early diagnosis. However, Diagnocat showed poor accuracy in assessing periodontal bone loss (κ = 0.19, AUC = 0.537), suggesting limitations in complex diagnostic cases. Conclusions: Diagnocat demonstrates potential as a supportive diagnostic tool for dental caries and missing teeth but requires refinement for periodontal assessment. Further research is necessary to improve algorithmic performance, adjust for intra-patient clustering effects, and optimize decision thresholds for better clinical applicability.