This review comprehensively examines the current status, clinical applicability, and future directions of artificial intelligence (AI) in diagnostic and laboratory dentistry. Dental AI primarily utilizes deep learning models based on radiographic and histopathological images to assist in diagnosis of conditions including dental caries, periodontal disease, periapical lesions, implant-related assessments, temporomandibular disorders, and oral squamous cell carcinoma (OSCC), enhancing diagnostic sensitivity and specificity. Recently, AI applications have expanded to salivary and blood biomarker analyses, integrating inflammatory cytokines (IL-1β, TNF-α, MMP-8), OSCC-related proteins (CYFRA 21-1, SCC-Ag, p53), salivary microbiome profiles, and blood-based indicators such as C-reactive protein, HbA1c, and bone metabolism markers, enabling predictive and personalized diagnostic modeling. The potential use of large language models (LLMs) has garnered attention, offering capabilities for analyzing electronic health records and clinical text data to support diagnosis, recommend treatment strategies, and assist in patient counseling and education. In the United States, several dental AI platforms, including Pearl Inc.’s Second Opinion®, have received FDA 510(k) clearance and are entering clinical practice, while in Korea, commercialization is progressing through Ministry of Food and Drug Safety approvals. Nevertheless, challenges remain, including insufficient data standardization, limited multi-institutional datasets, legal and ethical considerations, and integration with clinical workflows. To address these issues, multi-institutional prospective validation, development of generalizable models, multimodal AI research, and implementation of explainable AI are necessary. Overall, dental AI is evolving beyond image interpretation toward a multimodal clinical decision support system that integrates imaging, biomarkers, clinical information, and LLMs to support personalized diagnostics and treatment planning after validation.
Yeon–Hee Lee (Sun,) studied this question.