Comparative study of two software for the detection of cephalometric landmarks by artificial intelligence

Abstract

Manual, tedious cephalometric analyzes of a lack of productivity (errors in plotting and measurement) making the prospect of a fully automated algorithm turning out attractive. The objectives of the study were to evaluate the positioning (accuracy and precision) of common landmarks by two software programs offering detection by artificial intelligence (WebCeph™ and DentaliQ®) compared to a manual reference and then to the comparators. 68 lateral cephalograms were selected, 22 landmarks were distributed manually and then the differences between this gold standard and the points detected by each software were measured, as well as the success detection rate (SDR). Statistical analysis was carried out by “confidence ellipses” and two-tailed t-test (p-value of 5%). In terms of accuracy, WebCeph™ and DentaliQ® show a 2 mm SDR of 57.2% and 66.5% respectively. In terms of trueness, the best results are obtained for S, Na and the incisal edges. Large random errors are found for the points Po, So, ENA, ENP, Ba and Go. Other points like Pog and B show a large vertical dispersion. Overall, a slight advantage goes to DentaliQ® even if the difference is not significant. The detection precision still seems insufficient for an utilization without human supervision. The results are promising on the detection of certain points. The artificial intelligence saves time for the clinician but the change of positioning of points should still be possible. Advances are rapid and will probably be soon for an effective clinical use.