ABSTRACT This review scoped artificial intelligence (AI) in paediatric dentistry to summarize current applications and limitations. A comprehensive literature search was completed across PubMed, Google Scholar, Scopus, Embase, IEEE Xplore, and medRxiv/aArXiv. The preferred Reporting Items for Systematic Reviews and Meta‐Analyses Extension for Scoping Reviews (PRISMA‐ScR) checklist was used for reporting the present study. Studies concerning paediatric dentistry and utilizing AI, machine learning (ML) or large language models (LLMs) were included. Two independent reviewers screened and three extracted data on AI methodologies, performance metrics, and clinical applications. Fifty‐nine studies were included. Computer vision studies ( n = 32) most often used panoramic radiographs ( n = 14), mainly for tooth detection, numbering and segmentation ( n = 10) and caries detection ( n = 9). Non‐imaging ML studies ( n = 18) primarily used clinical variables and questionnaire data ( n = 8) or salivary/biological samples ( n = 10). Accuracy ranged from 50% to 100%, with sensitivity reaching 100% in biomarker‐based studies. LLM accuracy varied widely (33%–96%). Computer vision and non‐imaging ML show promise for paediatric dentistry, particularly in the detection of dental structures and diseases, as well as risk stratification. Variable LLM performance underscores the need for standardized evaluation before clinical integration. Clinical Relevance AI‐based tools may support paediatric dentists by improving tooth detection, dental structure assessment, caries detection, disease prediction and risk stratification. Computer vision and non‐imaging machine learning approaches appear to have potential, but their clinical value depends on robust external validation, transparent reporting and integration into paediatric dental workflows. Given their variable performance, AI tools should be used only as adjunctive tools or a second reviewer.
ABSTRACT Background Osteonecrosis of the jaw (ONJ) is a serious condition associated mainly with antiresorptive therapy and maxillofacial radiotherapy. As detection of its radiographic findings may facilitate patient referral, this study aimed to develop convolutional neural networks (CNNs) for ONJ screening in panoramic radiographs (OPGs). Methods A dataset of 888 ONJ‐positive OPGs and 888 ONJ‐negative controls was assembled and annotated using bounding boxes. The dataset was split at patient level 70–10–20 for training, validation, and testing. Multiple CNNs (ResNet18, ResNet34, MobileNetV2, EfficientNet‐B0) were trained on cropped image regions and evaluated using standard performance metrics. Performance on full OPGs (without cropping) was tested, and subgroup analysis was performed. Regions contributing to model predictions were visualised using Gradient‐weighted Class Activation Mapping (Grad‐CAM). Results EfficientNet‐B0 performed best, with per‐crop accuracy, sensitivity, specificity, precision, and F 1‐score of 92.24%, 89.80%, 94.69%, 94.42% and 82.05%, respectively. At 0.9 probability threshold, per‐image accuracy, sensitivity, specificity, precision, and F 1‐score equalled 76.65%. Grad‐CAM highlighted clinically relevant regions; however, differentiation from other changes in bone density was suboptimal. Conclusions Despite promising results at crop level, model performance in full OPG analysis is currently insufficient for clinical application. Further optimisation of localisation strategies and external validation are needed for reliable clinical application.
ABSTRACT Background Managing oral mucosal lesions in general dental practice is challenging, as visually similar conditions may require very different treatments. We developed an AI‐based triage system connecting intraoral image analysis with guideline‐based treatment recommendations, tested it externally, and compared its performance with clinicians' recommendations. Methods YOLOv8 and YOLOv11 were trained on 4079 intraoral photographs to classify 11 oral mucosal lesion types. These were mapped into five treatment tiers defined by urgency and management aim: T1 specialist referral for malignant and high‐risk lesions, T2 interdisciplinary referral for specialist‐led non‐cancer management, T3 risk‐adapted surveillance, T4 dental treatment with reassessment, and T5 no treatment. Model performance was compared with two general dentists and four specialists on an internal test set. YOLOv11 was additionally tested on an external dataset. Results YOLOv11 outperformed YOLOv8 in macro‐averaged precision, recall, and F1 score, achieving 0.72, 0.74, and 0.73 respectively. However, 25% of T1 high‐risk cases were under‐triaged. YOLOv11 significantly outperformed all clinicians, whose F1 scores ranged from 0.33 to 0.69. External performance decreased markedly (F1 score: 0.58), consistent with domain shift. Conclusion Guideline‐oriented AI triage is feasible but requires multicentre training for improved generalizability, as well as safety‐focused optimization, and prospective user evaluation before clinical use.
ABSTRACT Artificial intelligence (AI) is rapidly transforming clinical dentistry, with an increasing number of AI‐powered diagnostic and treatment planning tools entering the Australian healthcare market. As clinicians adopt AI‐based dental software, it is becoming increasingly important to understand the regulatory frameworks that govern them. This narrative review explores the regulatory approval pathways and governance of AI tools used in Australian dental and healthcare settings to inform and educate practitioners. Australia regulates AI‐based dental software primarily through the Therapeutic Goods Administration, which classifies eligible products as software as a medical device under a risk‐based, technology‐agnostic framework established by the Therapeutic Goods Act 1989. AI tools that function as clinical decision‐support systems are subject to mandatory conformity assessment and registration on the Australian Register of Therapeutic Goods (ARTG) prior to lawful supply. Although the Australian framework aligns with the International Medical Device Regulators Forum principles, the approval process is comparatively less transparent than that required in jurisdictions such as the United States and the European Union, and questions remain regarding the consistency of post‐market oversight. This review also highlights emerging regulatory challenges associated with adaptive AI systems. Software could initially receive approval as a low‐risk device that evolves to provide specific clinical advice beyond its approval scope. These updates to functionality risk bypassing established oversight pathways that rely on self‐reported intended use. Furthermore, currently available AI tools lack comprehensive public‐facing transparent evidence for regulatory compliance that is complicated by ambiguous and inconsistent terminology and adaptive AI system classification. Regardless of regulatory approval, clinicians remain professionally accountable for the use of AI tools in clinical decision‐making and for ensuring that patient care meets accepted standards of practice.
