Semester

Spring

Date of Graduation

2025

Document Type

Thesis

Degree Type

MS

College

School of Dentistry

Department

Orthodontics

Committee Chair

Khaled AlSharif

Committee Co-Chair

Peter Ngan

Committee Member

Peter Ngan

Committee Member

Guoqiang Guan

Abstract

Objective: To compare the accuracy and reliability of two commercially available Artificial Intelligent (AI) cephalogram tracing software programs to doctor traced cephalograms.

Materials and Methods: A sample of 100 pretreatment cephalograms were analyzed with AI based software programs, Angel Aligner iCeph and WebCeph, then compared to images traced by an orthodontist using Dolphin Imaging. Of the 100 samples, 50 were skeletally Class I and 50 were skeletally Class II. For each cephalogram produced, the American Board of Orthodontics 11 measurements were acquired from each of the three tracing programs. One way analysis of variance (ANOVA) followed by Tukey’s HSD test was conducted to evaluate the mean difference among three types of measurements for all patients and stratified subsamples of Class I and Class II patients. Pearson’s correlation coefficients were calculated to assess the intra-rater reliability. A two-sided p-value less than 0.05 was considered statistical significance.

Results: Pearson’s correlation coefficients for all variables were above 0.90, indicating an excellent test and retest reliability for the study. Compared to the doctor tracings, both AI software programs showed varying degrees of statistical difference. There were more statistical differences among angular measurements when compared to linear measurements. For Angel Aligner iCeph, this was primarily seen in the measurements FMA and L1 to MP. For Webceph, this was mainly seen in the measurements SNA, ANB, U1 to NA, and L1 to MP. Regarding the clinical significance (measurements differing 2 mm or 2 degrees), there was predominantly a difference in FMA, SNMP and L1 to MP for Angel Aligner iCeph. For Webceph, there was also a clinically significant difference found in L1 to MP.

Conclusion: Overall, other than FMA and L1 to MP, Angel Aligner iCeph produced results that were most in line with doctor traced cephalograms. Webceph, on the other hand, showed more issues with accuracy and reliability. It is important to recognize that both the Angel Aligner iCeph and WebCeph software programs have website pop-ups that notify the doctor that generated measurements and analyses are for reference only. It is advised to use both AI software programs with caution, remembering to utilize the semi-automatic adjustment tool to manually adjust the tracings as necessary to produce the most accurate analyses.

Download

Share

COinS