To assess the accuracy of digital models generated using commercially available software to predict anterior teeth root inclination characteristics and compare the results to relevant data obtained from CBCT images.
Materials and Methods:
Following sample size calculation and after evaluation of inclusion and exclusion criteria, pre-treatment maxillary and mandibular plaster models and the corresponding CBCT scans of 31 patients attending a private orthodontic clinic in Beirut, Lebanon, were randomly selected. The subjects represented both genders (males: 10; females: 21), ranging in age from 12 to 40 years. Plaster models were scanned using the high resolution of an Ortho Insight 3D ™ scanner (Motion View Software, Chattanooga, Tennessee, USA) and CBCT scans were taken using a Kodak 9500 Cone Beam 3D System (Carestream Health, Inc., Rochester, New York, USA). Elaboration of plaster models and CBCT data were performed according to specific technical steps. The Shapiro-Wilk test was used to test normality of measurements per tooth and crosstabulation to examine the independency between categorical variables. Statistical analysis was performed using χ2 test of association. If the measurements were II normally distributed, the t-test and ANOVA were used to examine two or more continuous independent variables. In cases of non-normality of measurements, the Mann-Whitney and Kruskal-Wallis tests for multi-comparison of continuous data were employed. To evaluate the reliability and consistency of the investigator, the paired t-test was used. P-value was set as ≤ 0.05 in all statistical analyses. To determine the intra-examiner error of the measurements, records of 5 random patients were selected and all measurements were repeated twice by the author after a two week interval and the correlations computed using intra-class correlation coefficient (ICC).
The maximum disparity in angle between images derived from digital models and CBCT data was almost 40 degrees (upper left canine). The disparity in upper right canines was similar. The upper and lower canines produced the worst results, followed by the lower lateral incisors. The upper central incisors showed the best results, although the maximum angle of difference exceeds 20 degrees (with the median only around 8 degrees). There was no statistically significant difference in measurements of difference between long axes of the root by using the two methods in each tooth (MDBLAR) according to Angle classification. Observation of the cases showed that the software frequently estimates angulations leading to overlapping of adjacent roots; a clinically unfeasible situation in the absence of extensive root resorption or root morphology variation. Errors were higher for canines and lower for central incisors. There was no correlation of errors between teeth.
Root morphology imaging prediction is not a primary function of this software and this study confirmed its limitation as a sole tool in routine clinical applications. At III present these predictions cannot be considered accurate or reliable unless correlated with a radiographic image.