Accurate identification of cephalometric landmarks is essential for orthodontic diagnosis and treatment planning. Manual landmarking is time-consuming, requires clinical expertise, and is susceptible to intra- and inter-examiner variability. Artificial intelligence (AI) has emerged as a potential tool to automate this process, improving efficiency and consistency.

To map the current evidence on the use of AI for cephalometric landmark detection, identify trends in the AI methodologies employed, and highlight gaps in the literature requiring further research.

This scoping review was conducted in accordance with the PRISMA-ScR guidelines. PubMed, EMBASE, and Medline were searched up to April 2024, and identified 68 eligible studies. Data extracted included datasets used, numbers of landmarks assessed and expert examiners involved, AI algorithms employed, and reported performance outcomes.

The included studies analysed approximately 450,000 lateral cephalograms, frequently using the IEEE ISBI Grand Challenge 2015 dataset. Convolutional neural networks (CNNs) were the most common AI architecture. Several AI systems achieved landmark localisation comparable to expert clinicians, with mean errors within 2 mm, although performance varied between studies and landmarks. Heterogeneity in study design, datasets, validation methods, and reporting standards limited comparison of the findings.

AI may have a role in supporting cephalometric landmark detection. However, considerable variation in reported performance outcomes was observed across studies, potentially reflecting differences in landmark identification protocols, algorithm design, and dataset characteristics. The evidence remains heterogeneous, and further research using standardised methodologies and diverse datasets is needed to evaluate the applicability of these systems in routine clinical practice.

DOI registration link on OSF: CGVSU (accessed on June 29, 2026).