Background:A variety of etiological variables, including systemic, local, and hereditary, can lead to tooth impaction, a common pathological dental disorder. The present study was conducted to evaluate use of artificial intelligence for detecting impacted third molars on cone-beam computed tomography scans.Materials & Methods:50 patients of both genders were selected. Monitoring for impaction, the number of impacted teeth, the number of teeth's roots and canals, and the teeth's relationship to nearby anatomical features (maxillary sinus and inferior alveolar canal) were compared between the AI application and the human observer. The Kappa analysis was used to assess the recorded parameters agreement between the deep-CNN-based AI application and the human observer.Results:AI detected impacted tooth rightly in 86% and tooth number in 95%. The difference was significant (P< 0.05). In maxilla, sinus relation was correctly detected in 91%, root number in 69%, canal number in 67%. In mandible, canal detected correctly in 92%, canal relation in 85%, root number detection in 84% and canal number detection in 65%. The difference was significant (P< 0.05). There was moderate agreement between moderate in Inferior alveolar canal/ lower impacted molar, good in maxillary sinus/upper impacted molar, moderate in root number accuracy, and low in canal number accuracy.Conclusion: High accuracy values in identifying impacted third molar teeth and their correlation with anatomical features were demonstrated by the CNN approach employed in this investigation. The diagnosis of dentomaxillofacial architecture and diseases can be improved with the application of additional algorithms and machine learning techniques, particularly for the detection of third molars.