Title

Comparison of manual and AI-driven cephalometric landmark identification: An observational study

Authors

Krishna P. Shah*, Pradeep Kawale, Snehal bhalerao, Siddhant D. Sarode, Mode Vaishali Vaijanath & Swapnali Sudhakar Khedkar

Affiliation

Department of Orthodontics and Dentofacial Orthopaedics, Yogita Dental College and Hospital, Khed, Maharashtra, India; *Corresponding author

Email

Krishna P. Shah - E-mail: shahkrishna.shah12@gmail.com

Pradeep Kawale - E-mail: dr.pradeepkawale@gmail.com

Snehal bhalerao - E-mail: snehalbhalerao1@gmail.com

Siddhant D. Sarode - E-mail: siddhantsarode18@gmail.com

Mode Vaishali Vaijanath - E-mail: vaishalimode02@gmail.com

Swapnali Sudhakar Khedkar - E-mail: swapnalikhedkar12@gmail.com

Article Type

Research Article

Date

Received July 1, 2026; Revised July 31, 2026; Accepted July 31, 2026, Published July 31, 2026

Abstract

Accurate cephalometric landmark identification is challenging because manual tracing is time-consuming, operator-dependent and prone to observer variability. Therefore, it is of interest to compare the AI-assisted and manual identification of 20 landmarks on 120 pre-treatment lateral cephalograms. Mean landmark error was 1.08 ± 0.46 mm manually and 1.34 ± 0.62 mm with AI, with lower AI accuracy for bilateral and low-contrast landmarks. AI significantly reduced tracing time to 0.74 ± 0.21 min versus 8.42 ± 1.36 min manually (p < 0.001). Thus, data shows that AI provides clinically acceptable accuracy with substantial time savings, although expert verification remains necessary for anatomically ambiguous landmarks.

Keywords

Artificial intelligence (AI), cephalometry, orthodontics, landmark identification, deep learning (DL), lateral cephalogram

Citation

Shah et al. Bioinformation 22(7): 4170-4175 (2026)

Edited by

Hiroj Bagde

ISSN

0973-2063

Publisher

Biomedical Informatics

License

This is an Open Access article which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. This is distributed under the terms of the Creative Commons Attribution License.