Assessment of Computerized Cephalometric Growth Prediction: A Comparison of Three Methods
thesisposted on 10.12.2012 by Matthew Sagun
In order to distinguish essays and pre-prints from academic theses, we have a separate category. These are often much longer text based documents than a paper.
The concept of cephalometric growth prediction has been around for decades, but it has only been readily available as in-office programs in the recent past. No objective attempts have been made to validate the accuracy of the growth prediction algorithms in these programs. This retrospective study compares the predictive ability of these currently available programs to provide a measuring stick for growth prediction. Hypothesis: The three algorithms (Bolton/Ricketts in Dolphin Imaging™ and Ricketts in RMODS® software, JOE CEPH®) tested in this study provide accurate growth predictions when compared to actual observed growth of untreated orthodontic subjects. Objective: To evaluate and compare the accuracy of three computerized growth prediction algorithms based on lateral cephalograms. Methods: Longitudinal sets of cephalograms from 56 untreated caucasian subjects were studied. Each subject had three cephalograms, with the first time point (T1) taken between the ages of 8 and 11 and subsequent cephalograms taken 2 years (T2) and 4 years (T3) after T1, that were all traced using Dolphin Imaging™. Two-year and four-year growth predictions were then performed using the Ricketts growth prediction algorithm (Alg 1) and the Bolton growth prediction algorithm in Dolphin Imaging™ (Alg 2) and the Ricketts growth prediction algorithm in RMODS® software, JOE CEPH® (Alg 3). The predictions were superimposed with their corresponding tracing of actual growth. Statistics were performed to determine the accuracy of the predictions as compared to actual growth using a clinical reference mean of 1.5 mm and to compare the algorithms to each other. Results: Alg 1 predicted 17 of 18 landmarks for the 2-year predictions for males, 18 of 18 landmarks for the 2-year predictions for females and 14 of 18 for the 4-year predictions for both males and females. Alg 2 predicted 18 of 18 landmarks for all 2-year predictions. For the 4-year predictions, Alg 2 predicted 15 of 18 and 12 of 18 using chronologic age for males and females, respectively. It also predicted 13 of 18 and 16 of 18 using skeletal age for males and females, respectively. Alg 3 predicted 18 of 18 landmarks for all the 2-year predictions. Alg 1 was the most different and Alg 2 and Alg 3 showed to be more similar. Conclusion: Overall, the three growth prediction algorithms tested provided accurate growth predictions when compared to observed growth of untreated orthodontic subjects.