Artificial intelligence web-based cephalometric analysis platform: comparison with the computer assisted cephalometric method

ABSTRACTPurpose The aim of this investigation was to evaluate the reliability and accuracy of cephalometric measurements of the web-based artificial intelligence cephalometric analysis platform in comparison with the computer assisted cephalometric analysis method.Materials and Methods 60 patients’ pretreatment lateral cephalograms were randomly selected. A total of 21 landmarks were identified by one operator and a total of 20 parameters were measured both AI based platform WebCeph® and Dolphin Imaging®. Measurements of AI landmarking were recorded. Then, the landmarks placed automatically by the AI (AI landmarking) were corrected manually (manual landmarking). All the measurements were recorded and performed once more after 4-weeks. Correlation between repeated measurements was evaluated by using the Pearson correlation coefficient. Paired t-test was used for comparisons between groups.Results Most of the measurements showed statistically significant differences between AI landmarking and manual landmarking 1, except for the angular measurements of the U1-SNº (P = 0.717), interinsizal angle (P = 0.410), and L1-NBº (P = 0,295). Most of the measurements were found to be statistically similar between manual landmarking 1 and manual landmarking 2, except for the angular measurement of the SN-GoGnº, IMPAº, linear measurements ANS-Me. The Pearson correlation coefficients of all cephalometric measurements were above 0.80.Conclusions All mean differences between the manual landmarking 1 and AI landmarking measurements were less than 2 degrees/2 mm, except for the nasolabial angle. Although WebCeph’s artificial intelligence algorithm is not sufficient to accurately determine the position of soft tissue landmarks, it becomes more suitable for clinical use with the control and manual correction of landmarks by observers.KEYWORDS: Artificial intelligenceautomatic landmarkingcephalometricWebCeph AcknowledgmentsWe thank Hatice Cansu Kış, PhD, from Gaziosmanpasa University (Tokat, Turkiye) for technical support and advice on statistical analyses.Disclosure statementNo potential conflict of interest was reported by the author.Authors contribution“HD made the cephalometric tracings, analysed and interpreted the cephalometric data, standardized the cephalograms and selected them according to the including criteria.Ethical approvalThe study was conducted according to the Declaration of Helsinki principles and was approved by the Scientific Research and Publication Ethics Committee at Nuh Naci Yazgan University (Approval No:1/393).