Artificial Intelligence in Dentistry: Applications in Diagnosis, Treatment Planning, and Clinical Decision-Making
Abstract
ABSTRACT
Artificial intelligence (AI), particularly deep learning, is rapidly entering dental practice. Convolutional neural networks can analyse radiographs, photographs, intraoral scans and cone-beam computed tomography (CBCT) images to detect and segment anatomical structures and pathology with accuracy that often matches, and sometimes exceeds, that of individual clinicians in controlled studies. Diagnostic applications include detection of caries on bitewing and periapical radiographs, assessment of periodontal bone loss and staging of periodontitis, identification of periapical lesions, screening for oral cancer and potentially malignant disorders, and automated cephalometric landmarking. AI also supports treatment planning: it can segment teeth and jaws from CBCT for implant planning and virtual orthodontic set-ups, design crowns and other restorations automatically, and predict outcomes such as tooth loss or treatment success. Several dental AI systems now have regulatory clearance as decision-support tools, and a randomised trial has shown that AI assistance can increase dentists' sensitivity for detecting caries. However, the evidence base has important limitations, including small, single-centre datasets, risk of bias, limited external validation, poor generalisability across populations and imaging devices, and few prospective studies measuring patient outcomes. Large language models have emerged as a new form of AI, with potential uses in documentation, patient communication and education, but they can generate inaccurate information. Ethical, legal and regulatory issues, including data privacy, bias, transparency, accountability and the risk of overtreatment, must be addressed. This review summarises current applications of AI in dental diagnosis, treatment planning and clinical decision-making, discusses their limitations, and outlines requirements for safe and effective adoption.
