Evaluation of reliability and reproducibility of artificial intelligence-based, computer-based and traditional manual tracing methods.

INTRODUCTION: Technological developments affect every aspect of life and science, including cephalometric evaluation methods in orthodontics. Artificial-intelligence-supported mobile Internet devices provide doctors the convenience of performing necessary analyses for patients anytime and anywhere. OBJECTIVE: This study aimed to compare the lateral cephalometric measurement methods of the Artificial Intelligence Tracking Method (AITM, WebCeph), Computer-Based Tracking Method (CBTM, NemoCeph), and Manual Tracking Method (MTM), to evaluate the reliability of medical technological approaches, compared to traditional methods. METHODS: Two hundred fifty lateral cephalometric radiographs were evaluated using three different methods: WebCeph, NemoCeph, and manual tracing. For this purpose, ten angular (SNA, SNB, ANB, Go-Gn/SN, Occlusal Plane/SN, U1/SN, U1/NA, L1/Mandibular Plane, L1/NB, and Interincisal angle [U1/L1]) and five linear (U1/NA, L1/NB, Wits Appraisal, Ls/E Plane, and Li/E Plane) parameters were measured, and the data were statistically analyzed. RESULTS: There was excellent reliability between the three methods in measurements of SNA, SNB, L1/NB Angle, L1/NB mm, ANB, Wits Appraisal, Ls/E Plane, Li/E Plane, Go-Gn/SN, and Occlusal Plane/SN (p<0.001 and ICC range 0.915 to 0.979). Regarding the agreement between the two measurements, there were statistically significant perfect agreements in all other parameters, except for two moderate ones. Statistically significant differences were found in almost all measured parameters between the tracing methods. CONCLUSION: There were significant differences in the majority of the measurement methods. Regarding reproducibility, the Interincisal Angle values showed low consistency with the AITM method, and the U1/SN values showed low consistency with both the CBTM and AITM methods.