A Benchmark Dataset for Automatic Cephalometric Landmark Detection and CVM Stage Classification

• Published in: Scientific data Vol. 12; no. 1; pp. 1336 - 13
• Main Authors: Khalid, Muhammad Anwaar, Zulfiqar, Kanwal, Bashir, Ulfat, Shaheen, Areeba, Iqbal, Rida, Rizwan, Zarnab, Rizwan, Ghina, Fraz, Muhammad Moazam
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 31.07.2025; Nature Publishing Group; Nature Portfolio
• Subjects: 692/53/2421; 692/698/3008; Anatomic Landmarks - diagnostic imaging; Annotations; Artificial intelligence; Automation; Benchmarking; Cephalometry; Cervical Vertebrae - diagnostic imaging; Cervical Vertebrae - growth & development; Classification; Data Descriptor; Datasets; Humanities and Social Sciences; Humans; Localization; Medical diagnosis; multidisciplinary; Orthodontics; Science; Science (multidisciplinary); Vertebrae
• ISSN: 2052-4463; 2052-4463
• DOI: 10.1038/s41597-025-05542-3

Accurate identification and localization of cephalometric landmarks are crucial for diagnosing and quantifying anatomical abnormalities in orthodontics. Traditional manual annotation of these landmarks on lateral cephalograms (LCRs) is time-consuming and subject to inter- and intra-expert variability. Attempts to develop automated landmark detection systems have persistently been made; however, they are inadequate for orthodontic applications due to the unavailability of a diverse dataset. In this work, we introduce a state-of-the-art cephalometric dataset designed to advance AI-driven quantitative morphometric analysis. Our dataset comprises 1,000 LCRs acquired from seven different imaging devices with varying resolutions, making it the most diverse and comprehensive collection to date. Each radiograph is meticulously annotated by clinical experts with 29 cephalometric landmarks, including the most extensive set of dental and soft tissue markers ever included in a public dataset. Additionally, we provide cervical vertebral maturation (CVM) stage annotations, marking the first standard resource for CVM classification. We anticipate that this dataset will serve as a benchmark for developing robust, automated landmark detection frameworks, with applications extending beyond orthodontics.