Imaging Sci Dent.  2025 Mar;55(1):11-21. 10.5624/isd.20240089.

Comparative accuracy of artificial intelligence-based AudaxCeph software, Dolphin software, and the manual technique for orthodontic landmark identification and tracing of lateral cephalograms

• Foroozandeh M
• Salemi F
• Shokri A
• Farhadian N
• Aeini N
• Hassanzadeh R

• KMID: 2566318
• DOI: http://doi.org/10.5624/isd.20240089

Abstract

Purpose
The aim of this study was to compare the accuracy of AI-based AudaxCeph software, Dolphin software, and the manual technique for identifying orthodontic landmarks and tracing lateral cephalograms.
Materials and Methods
In this cross-sectional study, 23 anatomical landmarks were identified on 60 randomly selected lateral cephalograms, and 5 dental indices, 4 skeletal indices, and 1 soft tissue index were measured. Each cephalogram was traced using 4 different methods: manually, with the Dolphin software, with the AudaxCeph software automatically, and with the AudaxCeph software in semi-automatic mode. The intra-class correlation coefficient (ICC) and Bland-Altman plots were used to evaluate the agreement between methods. Inter-observer and intra-observer agreements, calculated using the ICC, confirmed the accuracy, reliability, and reproducibility of the
results
.
Results
There was strong agreement among the AudexCeph (semi-automated or automated) AudaxCeph, Dolphin, and manual methods in measuring orthodontic indices, with ICC values consistently above 0.90. Bland-Altman plots confirmed satisfactory agreement between both versions of AudaxCeph (semi-automated and automated) with the manual method, with mean differences close to 0 and about 95% of data points within the limits of agreement. However, the semi-automated AudaxCeph showed greater agreement and precision than the automated version, as indicated by narrower limits of agreement. The ICC values for inter-observer and intra-observer agreements exceeded 0.98 and 0.99, respectively.
Conclusion
The semi-automated AudaxCeph software offers a robust and cost-effective solution for cephalometric analysis. Its high accuracy and affordability make it a compelling alternative to Dolphin software and the manual method.

Keyword

Artificial Intelligence; Deep Learning; Diagnostic Imaging; Cephalometry; Anatomic Landmarks; Orthodontics