(PR086) Comparison of AI-assisted Interpretation Versus Dental Students in the Diagnosis of Periapical Radiolucencies

Accurate diagnosis of periapical radiolucencies (PARL) is essential for timely intervention. Artificial intelligence (AI) has the potential to enhance radiographic diagnostic accuracy. This research aims to compare the effectiveness of Artificial Intelligence (AI) in detecting PARLs in periapical radiographs (PA) with predoctoral dental students. This study evaluated the sensitivity and specificity of AI detection models and compared them with interpretative skills of D3 and D4 dental students. FDA-cleared Overjet PARL-assist (OPA) AI module was used to detect and segment PARLs in 15 periapical radiographs. 47 D3 students and 42 D4 students interpreted these radiographs separately under standardized viewing conditions. The gold-standard for the lesions was established by a consensus of an OMR faculty and a calibrated Endodontic resident. The AI module had an 83% sensitivity and 96% specificity in detecting the PARL lesions. The dental students were able to identify PARLs in 69% of the cases. The students were unable to detect PARL 53% in Maxillary Molar PAs followed by 29% in Maxillary Anterior PAs, 15% in Mandibular Molar PAs, 10% in Maxillary Premolar PAs and 2% in Mandibular Premolar PAs. Out of 23 PARL lesions, the AI module missed 4 PARLs. The sensitivity of the AI module was augmented to 91% with a student assisted AI-diagnosis. Augmented Intelligence (AuI) increased AI sensitivity by 8%. Students were more likely to miss maxillary molars and anterior PARLs. AI has the potential to enhance diagnostic accuracy and efficiency, while still relying on human judgement for final clinical decision-making.