Comparison of linear and angular cephalometric measurements in CBCT using semi-automated software and artificial intelligence-generated cephalograms.

BACKGROUND

This study aimed to assess the reliability of cephalometric measurements using CBCTs with two semi-automated software programs (InVivoDental and Romexis) and an artificial intelligence-based platform (Diagnocat).

METHODS

This cross-sectional reliability study retrospectively analyzed a total of 29 CBCT scans (15 from Northern Cyprus and 14 from Egypt). Eleven measurements were included to assess vertical relationship, and anteroposterior relationship, using two software programs and an AI-based website.

RESULTS

Intraobserver reliability, assessed through Bland-Altman plots, demonstrated high consistency. Interobserver agreement showed excellent reliability for most measurements, with ICC values exceeding 0.90 for both InVivoDental and Romexis software. Most measurements did not show significant differences between the two semi-automated programs and the AI-based algorithms. However, the AI system indicated lower measurements for Wits Appraisal and U1-SN angle compared to the two software systems.

CONCLUSION

In conclusion, this study underscores the importance of evaluating the reliability of cephalometric measurements on CBCTs, particularly when utilizing semi-automated software programs and artificial intelligence-based platforms. Our analysis revealed consistent intra-observer repeatability across most measurements. However, it is noteworthy that certain metrics exhibited reduced repeatability when assessed using AI-generated cephalograms, suggesting potential areas for improvement in AI algorithms.