A primer and overview of the role of artificial intelligence in oral and maxillofacial radiology

The purpose of this introductory article is to serve as a primer and overview of the role of artificial intelligence (AI) in oral and maxillofacial radiology (OMR). Included is a summary of the goals of artificial intelligence in research and practice in OMR and general dentistry. Of primary importance is the fact that AI research should be directed toward the improvement of clinician diagnostic performance. While standalone testing is important, the goal should always be to improve clinician patient care. Fundamental basics of AI are addressed in terms of how deep learning (DL) models are created, including a description of the basic methodologies for training, validation, and testing. A discussion regarding reference standards for determining “ground truth”, such as consensus panels, biopsy results, and cone beam CT as a standard for lesions on two dimensional radiographs is provided. An overview survey of the AI current literature for OMR is provided, highlighting representative papers from the various types of diagnostic tasks in imaging. The article concludes with an assessment of the current role of AI in OMR practice and research, and the way forward. One of the key fundamental factors that must be kept in mind is that AI detects but does not diagnose. Diagnosis is the responsibility of the clinician, so caution is required when reviewing outcomes from AI models based on DL.