Accuracy and Reproducibility of Semidigital Versus Fully Digital Cephalometric Tracings Using a New Computer Program Versus Conventional Methods (Gold Standards): A Preliminary Study.

Abstract

Introduction: Cephalometric tracing can be done either conventionally or using computers. Digital dentistry and digital orthodontics have considerably facilitated procedures. Still, their diagnostic accuracy needs assessment. Many orthodontic programs have been developed for this purpose. The efficacy and reliability of such software are usually compared with the conventional method (gold standard). We used novel and more stringent methods of assessment to test a program in this regard. Methods: This study was performed on 10,302 tracing evaluations within 101 cases. Lateral cephalograms of 101 patients were landmarked using two methods (on paper vs. on a computer screen) and traced using three methods (completely conventionally [gold standard]; landmarks were identified on paper, but measurements were calculated by computer; landmarks were identified on the computer screen, and measurements were calculated by the computer program). A total of 15 landmarks and 17 cephalometric tracing measurements were determined via the abovementioned methods. The tracing errors were defined as differences between each pair of tracing methods, as well their absolute values (a total of 6 different tracing errors). Intraclass correlations were calculated for tracing values. Tracing errors were compared with the value 2, as the clinically acceptable range. However, they were also compared with the values zero as well as one hundredth of the mean of gold standard (as a more conservative value), using a one-sample t-test (α = 0.05). Results: All tracing errors were smaller than the clinically acceptable limits. Moreover, most simple errors were close to zero, and/or below the criterion of 1/100 of the mean of the gold standard. Furthermore, the more difficult error tests, that is, the directionless absolute errors, were all below 2; additionally, they were either below the 1/100 of absolute of the gold standard means or at the level of those means. Finally, the intraobserver reliabilities were high. All the 102 simple errors and absolute errors (on 101 lateral cephalograms) were significantly below 2 (p < 0.0005, clinically acceptable). Conclusions: The accuracy was appropriate. Of the 51 simple tracing errors, only 9 were significantly greater than zero, and many of them were below or at the level of 1/100 of the gold standard means. All the directionless (absolute) errors were significantly greater than zero. However, in the case of those calculated as "absolute value of (gold standard subtracted by fully digital method)," all errors were below or at the level of 1/100 of the absolute of gold standards' means. The intraobserver reliabilities were high.