Evaluating the Accuracy of Automated Cephalometric Analysis Based on Artificial Intelligence

Objective: To determine the correlation and comparison between artificial intelligence (AI) cephalometric tracing and manual human tracing of cephalograms. Methodology: This cross sectional comparative study was conducted on cephalograms from 100 participants at Saidu College of Dentistry, Swat, using a non-probability sampling technique. Informed consent was not required as the study utilized existing radiographic records with prior patient consent. Age and gender were documented, followed by the manual tracing of 100 cephalograms by a sole operator, which were subsequently analyzed using AudaxCeph software. Pearson correlation testing and paired t-tests were utilized for analysis. Results: The participants' average age was 19.39 years (SD ± 6.24), with a gender distribution of 56% female and 44% male. The results revealed strong positive correlations (0.91 to 0.98) between manually and software-traced cephalometric parameters, with statistically significant p-values (<0.01). Paired t-tests showed no significant differences across various parameters, affirming the reliability of the software-generated cephalometric measurements. Conclusion: This research showcases the precision of the software in evaluating distinct cephalometric measurements, emphasizing its promise as a dependable instrument for orthodontic diagnosis and planning treatments.