Anterior teeth root morphology prediction as derived from digital models: A comparative study of plaster study casts and natural teeth from dry skulls

Abstract

Aim: To assess the accuracy and reliability of the digital images of dental models generated by specific commercially available software in predicting root morphology characteristics in comparison to the corresponding natural teeth.

Materials and Methods: The initial sample of 56 natural teeth derived from14 dry human skulls with permanent dentition. From the 14 dry skulls, there were 11 mandibles and 3 maxillae. The inclusion criteria used in this study were that all the teeth should derive from permanent dentition of adults and should present normal crown morphology. Primary teeth, teeth with abnormal tooth morphology and restorations, and jaws resembling craniofacial anomalies or syndromes were excluded. Based on these criteria, one maxillary right lateral incisor was excluded since it had a fractured crown. Therefore, the total number of the sample of teeth used was 55. The sample size calculation required 43 teeth. Upper and lower dental arch impressions, utilizing alginate material, were taken and plaster study models were fabricated. The maxillary and mandibular plaster models were ii scanned using the Ortho Insight 3D laser scanner. Using the Ortho Insight 3D software (version 4.0.6), measurements, to the nearest 0.01 mm, were recorded by locating relevant landmarks using the intrinsic linear measurement function of the software. Based on these landmarks, the software produced virtual roots for the selected teeth. Then both the scanned casts and their root predictions were exported in stereolithography (STL) files. All corresponding 55 natural teeth were removed from the upper and lower jaws of the dry skulls and scanned using the Identica extraoral white-light scanner in order to calculate their actual root morphological characteristics. In order to acquire the whole tooth geometry, the teeth were scanned twice, once to acquire the crown and the cervical part of the root, and a second time to acquire the remaining part of the root, including the apex. The two scanned segments were joined in software by superimposing them along their common part. Finally, the accuracy of the digital models generated by the Ortho Insight 3D laser scanner in predicting root morphology characteristics was assessed by comparing these results to the corresponding measurements of the 55 natural teeth. The long axes of the tooth models obtained from the software prediction and the scanning of the actual teeth were computed and the discrepancy between them evaluated.

Results: The error of the method was evaluated by repeating the measurements on 14 teeth. The differences in root mean squared error (RMSE) of superimposition of the two scanned segments (crown - root) of each tooth ranged from -0.006 to 0.010 mm. The differences in RMSE of superimposition of each tooth on its corresponding cast ranged from -0.040 to 0.057 mm. Both these error ranges showed acceptable accuracy. The error in calculating the angulation between the predicted and actual long axes ranged from -3.2 to 2.3 degrees (95% limits of agreement: -3.7 to 3.2). iii The predicted tooth angulation was found to differ significantly from the actual angulation, both statistically and clinically. The angle between the predicted and actual long axes ranged from 2.0 to 37.6 degrees (average: 9.7 degrees; median: 7.4 degrees).

Conclusions: Based on the results of this study, the degree of error is higher for mandibular central incisors and maxillary lateral incisors, and lower for maxillary central incisors and mandibular lateral incisors. Further investigations and improvements of the software are needed before it can be considered clinically effective.