Compatibility of the superimpositions of cephalometric radiographs made by artificial intelligence software

AIM: The aim of this study was to evaluate the compatibility of superimpositions of before and after orthodontic treatment, made with digital cephalometric tracing programs (Facad and WebCeph) that provide automatic detection of landmarks by using fully automatic artificial intelligence software, with semi-automatic software program (Dolphin Imaging) superimpositions in which landmarks are determined manually. MATERIALS AND METHOD: Lateral cephalometric radiographs of 20 patients were taken one year apart. While 20 cephalometric landmarks were determined by the researcher in the Dolphin program, they were determined by artificial intelligence in the Facad and WebCeph programs. Sella-Nasion and Porion-Orbitale planes were taken as reference for cephalometric superimpositions with three software programs. The distances between the reference points to the X and Y coordinates were measured on the Image-J program. Friedman test, Wilcoxon sign test and Intraclass Correlation Coefficient were used for statistical analysis. RESULTS: In the measurements made on the SN and POr planes; there wasn't statistically significant difference between the measurements made with Dolphin, Facad and WebCeph methods at all points (p>0,05). In the measurements made in X and Y coordinates a positive directional, moderate and good compatibility was found; between Dolphin and Facad programs, at 10 points in the SN plane and 34 points in the POr plane; between Dolphin and WebCeph programs, 13 points in the SN plane, 1 point in the POr plane; between Facad and WebCeph programs at 5 points on the SN plane and 4 points on the POr plane (p≤0,05). CONCLUSION: Although there isn't statistical difference in the distances of the superimpositions to the X and Y coordinates made with Facad and WebCeph artificial intelligence programs compared to the manual measurements made using the Dolphin program; The Facad program showed significant compatibility with Dolphin cephalometric registrations in both SN and POr planes compared to WebCeph. Keywords: Cephalometry, Artificial Intelligence, Superimposition