A scoping review evaluating artificial intelligence in dental imaging and forensic odontology shows that deep learning models achieve high diagnostic accuracy, though methodological constraints mean these tools should currently assist rather than replace specialists. Published in BMC Medical Imaging, the review notes that growing interest in artificial intelligence stems from improved objectivity, efficiency, and repeatability in forensic analysis. However, the authors conclude that automated algorithms should currently serve as decision-support tools with humans remaining in the loop.
Authors Gurbet Alev Oztas Sahiner and Bengisu Karatas conducted the review to map artificial intelligence applications in forensic odontology. Their investigation assessed primary forensic tasks, including dental age estimation, biological sex estimation, human identification, and tooth classification. The researchers also examined recent multimodal artificial intelligence applications in forensic dentistry, while detailing current diagnostic performance, existing methodological limitations, and important directions for future investigation.
The scoping review was carried out according to the PRISMA-ScR guidelines. The researchers systematically searched three major electronic literature databases: PubMed/MEDLINE, Scopus, and Google Scholar for relevant studies published from January 2022 to December 2025. The search strategy combined keywords related to the main topic with explicit inclusion and exclusion criteria to identify original research across the field.
The review included original studies evaluating machine learning, deep learning, hybrid models, and multimodal large language models applied to forensic dentistry. Two independent reviewers performed duplicate study selection, data extraction, and formal methodological quality assessment using the QUADAS-AI tool. Because the selected literature exhibited pronounced methodological heterogeneity, the authors conducted a narrative synthesis rather than a quantitative statistical meta-analysis.
A total of twenty articles met all predefined criteria for inclusion. Among the analyzed applications, dental age estimation was the most widely studied forensic problem. Human identification constituted the second most common subject of study, followed in descending frequency by biological sex estimation and tooth classification.
Deep learning architectures based on convolutional neural networks demonstrated high diagnostic accuracy across various dental imaging modalities. The evaluated networks performed effectively on panoramic radiographs, cone-beam computed tomography images, intraoral scans, and cephalometric radiographs. These results show that deep learning systems can process diverse two-dimensional and three-dimensional radiographic records commonly utilized in dental forensic investigations.
Despite these positive diagnostic results, the evidence base revealed notable methodological weaknesses. The researchers emphasized that existing literature was limited by retrospective study designs, reliance on patient-specific sample sets, a widespread lack of external validation, and high methodological heterogeneity. As a result, although artificial intelligence offers strong potential to improve standardization, efficiency, and objectivity in forensic analysis, the authors stress that it must presently function solely as a decision-support aid with human oversight.
Subject of Research: Applications and methodological limitations of artificial intelligence in forensic odontology and dental imaging
Article Title: Artificial intelligence in dental imaging and forensic odontology: clinical applications, performance, and challenges – a scoping review
Article References: Oztas Sahiner, G. A., & Karatas, B. (2026). Artificial intelligence in dental imaging and forensic odontology: clinical applications, performance, and challenges – a scoping review. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-03073-w
Image Credits: AI Generated
DOI: 10.1186/s12880-026-03073-w
Keywords: artificial intelligence, forensic odontology, dental imaging, scoping review, dental age estimation, biological sex estimation, human identification, tooth classification, convolutional neural networks, deep learning, machine learning, multimodal large language models, panoramic radiographs, cone-beam computed tomography, intraoral scans, cephalometric radiographs, QUADAS-AI, PRISMA-ScR

