Diagnostic Accuracy of AI in Radiographic Detection of Dental Caries and Periapical Lesions

Abstract

Background: Accurate detection of dental caries and periapical lesions is critical for timely intervention and preservation of tooth structure, yet conventional radiographic interpretation is limited by observer variability and diagnostic fatigue. Recent advances in artificial intelligence (AI) offer automated image analysis with the potential to enhance diagnostic consistency and sensitivity in dental radiology. Objective: To compare the diagnostic accuracy of an AI-based radiographic tool with conventional clinical and radiographic examination for detecting dental caries and periapical lesions in adult dental patients. Methods: In this prospective diagnostic accuracy study, 240 adults undergoing intraoral periapical and bitewing radiography at a tertiary dental hospital in Lahore were consecutively enrolled. Two calibrated dentists performed conventional examinations using ICDAS II and PAI, blinded to AI outputs. A deep learning–based AI software analyzed all radiographs. Expert consensus by a radiologist and endodontist served as reference standard. Sensitivity, specificity, predictive values, accuracy, and area under the ROC curve (AUC) were calculated; McNemar’s and DeLong’s tests compared methods. Results: For caries detection, AI achieved sensitivity 91.7%, specificity 89.2%, and AUC 0.94, versus 83.4%, 81.6%, and 0.87 for conventional examination (all p ≤ 0.001). For periapical lesions, AI sensitivity, specificity, and AUC were 93.5%, 88.9%, and 0.96, compared with 84.8%, 80.2%, and 0.85 for conventional methods (all p ≤ 0.002). Conclusion: AI-based radiographic analysis demonstrated significantly superior diagnostic accuracy to conventional examination for both dental caries and periapical lesions, supporting its use as an adjunctive tool in routine dental diagnostics.