Effect of Artiﬁcial Intelligence versus guided Landmarks identification on the accuracy of the Lateral Cephalometric Analysis

Hemeda, Mohamed G.; Refai, Wael M.; Elshal, Mohamed G.; Mohammed, Mostafa A.; Abdelhameed, Ahmed N.

doi:10.21608/eos.2021.104517.1032

Effect of Artiﬁcial Intelligence versus guided Landmarks identification on the accuracy of the Lateral Cephalometric Analysis

Document Type : Original Article

Authors

¹ Candidate of orthodontics, Department of orthodontics, Faculty of Dentistry, Minia University, Minia, Egypt.

² Professor of orthodontics, Faculty of Dentistry, Minia University, Minia, Egypt.

³ Lecturer of orthodontics, Faculty of Dentistry, Minia University, Minia, Egypt.

10.21608/eos.2021.104517.1032

Abstract

Objective: This study aimed to compare the effect of Artiﬁcial Intelligence versus guided Landmarks identification on the accuracy of the Lateral Cephalometric Analysis.
Methodology: Three orthodontic specialists identified 17 radiographic landmarks manually for 22 different types of angular and linear measurements of 50 lateral cephalometric radiographs then tracing and analysis were done by Artificial Intelligence based software (Webceph) and Automated cephalometric analysis software (Romexis software). The measurements of the two softwares compared to humans’ gold standard (Mean values of the three examiners).
Results: comparison between humans’ gold standard and (Wepceph) the AI’s predictions showed no proportional bias in 12 parameters, The mean differences range from 0.2° to 2.9° for angular measurements except Gonial angle 4.55°, Upper 1 to NA angle 3.78° and IMPA 3.72° and from 0.25 to 1.67 mm for linear measurements. Comparison between humans’ gold standard and Automated cephalometric analysis software (Romexis software) showed proportional bias in 19 parameters, The mean differences range from 0.16° to 12.67° for angular measurments and from 1.01 to 13.38 mm for linear measurements.
Conclusions: AI based software is able to identify landmarks of cephalometric X-rays at almost the same quality level as experienced human examiners (current gold standard).comparison between the two types of softwares showed that the accuracy of AI based (Webceph) software is better than the automated cephalometric analysis (Romexis) software.

Keywords