AI Accuracy in Detecting Cephalometric Points
Author Information
Author(s): Ribas-Sabartés Júlia, Sánchez-Molins Meritxell, d’Oliveira Nuno Gustavo, Yang Yanqi
Primary Institution: Departamento de Odontoestomatología, Facultad de Medicina y Ciencias de la Salud, Universidad de Barcelona
Hypothesis
What forms of AI have the lowest error and highest reproducibility in detecting cephalometric landmarks in lateral cranial teleradiography compared to professionals?
Conclusion
CNN-based AI algorithms are the most effective for detecting cephalometric points, but they still do not match the accuracy of orthodontists.
Supporting Evidence
- CNN-based AI algorithms showed point localization accuracy ranging from 64.3% to 97.3%.
- Manual localization of points remains more accurate than automated methods.
- AI demonstrates better reproducibility compared to detection performed by professionals.
Takeaway
This study looked at how well AI can find important points in X-ray images of the head. It found that while AI is fast and consistent, it still isn't as accurate as a human orthodontist.
Methodology
A systematic review of studies from 2013 to 2023 assessing AI algorithms for detecting cephalometric landmarks in 2D radiographs.
Potential Biases
Six studies were classified as high-risk and seven as low-risk based on bias assessment tools.
Limitations
The review excluded studies requiring advanced AI knowledge and those involving patients with anomalies, which may limit the generalizability of findings.
Participant Demographics
Studies included various demographics, but specific details on age and gender were not consistently reported.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website