Surface analysis of study models generated from OrthoCAD and cone-beam computed tomography imaging

ORIGINAL ARTICLE Surface analysis of study models generated from OrthoCAD and cone-beam computed tomography imaging Kurtis G. Lightheart, a Jeryl D. English, b Chung H. Kau, c Sercan Akyalcin, d Harry I. Bussa, Jr, e Kathleen R. McGrory, f and Kevin J. McGrory f Houston, Tex, and Birmingham, Ala Introduction: The purpose of this research was to determine the accuracy of digital models generated by conebeam computed tomography and compare it with that of OrthoCAD models (Cadent, Carlstadt, NJ) for orthodontic diagnosis and treatment planning by using surface area analysis. Materials: Two sets of maxillary and mandibular digital models of 30 subjects were obtained. The models were made from impressions scanned with OrthoCAD and by conversion of related cone-beam computed tomography files. Each patient s matched pairs of maxillary and mandibular models were superimposed by using a software program and a best-fit algorithm; surface-to-surface analysis was then performed. The average linear differences between the 2 files at all points on the surfaces were measured, and tolerance levels of 0.25, 0.5, 0.75, 1.0, 1.25, and 1.5 mm were set to determine the surface correlation amounts between the 2 files. Additionally, 6 linear measurements from predetermined landmarks were also measured and analyzed. Results: The average maxillary model linear difference was 0.28 to 0.60 mm, whereas the average mandibular model linear difference ranged between 0.34 and 0.61 mm. Greater than a 90% surface correlation was obtained on average at 1.00 mm in the maxillary models and at 1.25 mm in the mandibular models. The mean differences obtained from the linear measurements of the maxillary and mandibular models were 0.071 and 0.018 mm, respectively. Conclusions: Surface-to-surface analysis of OrthoCAD and digital models generated by cone-beam computed tomography pointed to a fair overlap between the protocols. The accuracy of digital models generated by cone-beam computed tomography is adequate for initial diagnosis and treatment planning in orthodontics. (Am J Orthod Dentofacial Orthop 2012;141:686-93) Acomprehensive and accurate treatment plan can be possible only by a thorough evaluation of dental models, intraoral and extraoral photographs, and panoramic and cephalometric images. 1 a Resident, Department of Orthodontics, School of Dentistry, University of Texas Health Science Center, Houston. b Chair and professor, Department of Orthodontics, School of Dentistry, University of Texas Health Science Center, Houston. c Chair and professor, Department of Orthodontics, School of Dentistry, University of Alabama, Birmingham. d Assistant professor, Department of Orthodontics, School of Dentistry, University of Texas Health Science Center, Houston. e Clinical professor, Department of Orthodontics, School of Dentistry, University of Texas Health Science Center, Houston. f Clinical assistant professor, Department of Orthodontics, School of Dentistry, University of Texas Health Science Center, Houston. The authors report no commercial, proprietary, or financial interest in the products or companies described in this article. Reprint requests to: Sercan Akyalcin, University of Texas Health Science Center, School of Dentistry, Department of Orthodontics, 6516 MD Anderson Blvd, Suite 371, Houston, TX 77030; e-mail, sercan.akyalcin@uth.tmc.edu. Submitted, July 2011; revised and accepted, December 2011. 0889-5406/$36.00 Copyright Ó 2012 by the American Association of Orthodontists. doi:10.1016/j.ajodo.2011.12.019 Recent advances in digital technology have vastly improved the diagnostic phase of orthodontic treatment, and analog records have quickly been replaced by digital formats. 2,3 The digital alternative offers a great advantage when contemplating the storage space required for traditional plaster models. Digital models can be stored virtually instead of physically, thus reducing the need for storage space along with reducing the costs involved. 4-8 Advances in technology and software have allowed digital models to be generated from cone-beam computerized tomography (CBCT) scans. Acquired digital imaging and communications in medicine (DICOM) files captured by the orthodontist can be uploaded to a company s Web site, and a digital model can be obtained from the file. Once the model is generated by the proprietary software, the orthodontist can download, view, manipulate, and evaluate the models using software provided by the company. This technology allows for all anatomic structures from the CBCT image captured during the scan to be viewed. For example, 686

Lightheart et al 687 Fig 1. Superimposition of OrthoCAD and CBCT-generated models. roots of teeth, temporomandibular joint structures, impacted teeth, and bone levels are all visible. The practitioner can view, evaluate, and manipulate the models and bases alone without any accompanying bony structures. 9 However, the disadvantages to this approach are the lack of gingival tissues in the digital models and concern about possible unnecessary overexposure to radiation. Cone-beam technology is at the forefront of digital radiography, and generating digital models for diagnostic purposes from this information is a potential advancement that will help in the digitization of the orthodontic office. 10 However, digital models offer an advantage over plaster models only if they are proven to be as accurate for the diagnostic phase of treatment. A recent study showed that space analysis on OrthoCAD digital models (Cadent, Carlstadt, NJ) is clinically acceptable and reproducible when compared with traditional plaster study model analyses. 11 Fleming et al 12 concluded that, overall, the absolute mean differences between direct and indirect measurements on plaster and digital models were minor and clinically insignificant. The accuracy from linear measurements obtained from study casts generated by CBCT was also considered adequate for initial diagnosis and treatment planning in orthodontics, as shown elsewhere. 13,14 Validation of surface characteristics of digital models might particularly be important when considering various appliance designs used in clinical applications. The purpose of this study, therefore, was to examine the accuracy of digital models generated by CBCT compared with OrthoCAD digital models by using surface area shell analysis of stereolithography files. MATERIAL AND METHODS Thirty subjects were included in the study. Two sets of maxillary and mandibular digital models were obtained from them by scanning impressions with OrthoCAD and conversion of their CBCT files with Anatomage software (Anatomage, San Jose, Calif). All subjects were in the permanent dentition and had their impressions and baseline CBCT images obtained at the time of initial records for diagnosis and treatment planning at the Department of Orthodontics, School of Dentistry, University of Texas Health Science Center at Houston. A Galileos Comfort x-ray unit (Sirona Dental Systems, Bensheim, Germany) was used to capture the CBCT images. With 200 exposures from a 14-second cycle in a 220 segment, a high level of detail from a 15 3 15 3 15-cm volume capture image, and a large dental volume ranging from the tip of the nose to the chin and the mandibular joints, the cone-shaped conebeam radiation beams of the Galileos x-ray detector allowed for small-region close-up views at double the detail without an additional scan. Image reconstruction time was approximately 4.5 minutes, and voxel size was 0.30 mm. The CBCT images were uploaded to Anatomage and digitally reformatted to include volume rendering and image conversion. After this conversion, the 3- dimensional digital models were available for downloading from the server. American Journal of Orthodontics and Dentofacial Orthopedics June 2012 Vol 141 Issue 6

688 Lightheart et al Fig 2. Surface correlation analysis at 6 tolerance levels: a, 0.25 mm; b, 0.5 mm; c, 0.75 mm; d, 1.00 mm; e, 1.25 mm; f, 1.5 mm. The OrthoCAD models were obtained from the maxillary and mandibular impressions taken with Identic alginate (Dux Dental, Oxnard, Calif), along with bite registrations taken with polyvinylsiloxane material (Blue Moose, Parkell, Edgewood, NY). After disinfection, the records were placed in sealed plastic bags and mailed to OrthoCAD. Once received, poured, and digitized, the OrthoCAD digital models were available for download from the company s Web site. Stereolithography files were obtained from both the CBCT and the scanned impressions. This is a file format native to the stereolithography computer-aided design software created by 3D Systems (Rockhill, SC). In this study, Stereolithography files were required for both digital formats because the surface shapes and volumes were analyzed. Stereolithography files or shells were opened in the Rapidform software program (Inus Technology, Seoul, South Korea) for surface-to-surface analysis. In Rapidform, the stereolithography files from the CBCT were cropped to remove hard tissue, leaving as much tooth structure as possible above the alveolar bone crest. The OrthoCAD stereolithography files were also cropped to remove as much gingival tissue as possible. The cropped stereolithography files from both the CBCT scans and OrthoCAD were then superimposed, June 2012 Vol 141 Issue 6 American Journal of Orthodontics and Dentofacial Orthopedics

Lightheart et al 689 Fig 3. Linear measurements on digital models. and the Rapidform software obtained a best-fit from the 2 files (Fig 1). Once superimposed, a surfaceto-surface analysis was performed to compare the accuracy of the CBCT-generated digital models with the OrthoCAD digital models. This analysis averaged the linear differences between the 2 files at all points on the surfaces. For this study, tolerance levels were set to visualize the discrepancies between the 2 files that fell outside the set tolerance levels. The tolerance levels used in the study were 0.25, 0.5, 0.75, 1.0, 1.25, and 1.5 mm (Fig 2). During the visualization process, a linear difference less than the tolerance level appeared black on the superimposed image, and a linear difference greater than the tolerance level appeared as a color on a scale depending on the amount of the difference. Thus, black on the scale indicated that the linear difference fell within the tolerance range, and the colored regions on the scale indicated that the linear difference fell outside the tolerance level; blue was the most correlated, and red was the least. Mean differences at each tolerance level were recorded and compiled. In addition, the total surface areas corresponding to the tolerance levels were recorded as percentages. Six linear measurements (Fig 3) were made on each maxillary and mandibular model. The digital models by Anatomage were measured by using the proprietary software, InVivoDental, and the OrthoCAD models were measured by using the OrthoCAD proprietary software, OrthoCAD. Definitions of the linear measurements are summarized in Table I. Statistical analysis Intraclass correlation coefficients were calculated for the measurements obtained from maxillary and mandibular OrthoCAD and CBCT-generated models to test the reliability of the CBCT-generated models. IBM SPSS Statistics (version 19; IBM, Armonk, NY) was used to analyze the data. To further graphically display the levels of error between the 2 techniques, Bland-Altman analysis was performed by using XLSTAT Mac (version 2011; Addinsoft, New York, NY). RESULTS The average maxillary linear differences between the CBCT-generated and the OrthoCAD models were 0.28 to 0.60 mm, with a median of 0.36 mm. The average mandibular linear difference range between the CBCTgenerated and the OrthoCAD models was 0.34 to 0.61 mm, with a median of 0.44 mm (Table II). The results of the surface overlap for the maxillary and mandibular models in each tolerance level are presented in Table III. The average overlap of the surfaces for the maxillary models ranged from 45.76% at 0.25 mm of tolerance to 99.59% at 1.50 mm of tolerance. The average for the American Journal of Orthodontics and Dentofacial Orthopedics June 2012 Vol 141 Issue 6

690 Lightheart et al Table I. Definitions of linear measurements on the digital casts Maxillary cast measurements UR4-UL4 UR5-UL5 UR4-ML UR5-ML UL4-ML UL5-ML Mandibular cast measurements LR4-LL4 LR5-LL5 LR4-ML LR5-ML LL4-ML LL5-ML Buccal cusp tip of the upper right first premolar to the same point on the upper left first premolar Buccal cusp tip of the upper right second premolar to the same point on the upper left second premolar Buccal cusp tip of the upper right first premolar to the maxillary midline Buccal cusp tip of the upper right second premolar to the maxillary midline Buccal cusp tip of the upper left first premolar to the maxillary midline Buccal cusp tip of the upper left second premolar to the maxillary midline Buccal cusp tip of the lower right first premolar to the same point on the lower left first premolar Buccal cusp tip of the lower right second premolar to the same point on the lower left second premolar Buccal cusp tip of the lower right first premolar to the mandibular midline Buccal cusp tip of the lower right second premolar to mandibular midline Buccal cusp tip of the lower left first premolar to the mandibular midline Buccal cusp tip of the lower left second premolar to mandibular midline Table II. Maxillary and mandibular surface-to-surface analysis of CBCT-generated and OrthoCAD model overlays Maxillary casts Average distance (mm) Mandibular casts mandibular models ranged from 42.51% at 0.25 mm to 94.85% at 1.50 mm. Table IV displays the means and standard deviations of the linear measurements on both the OrthoCAD and the CBCT-generated models. These results indicate strong agreement (0.87 #single measure ICCs #0.99) between the 2 methods based on the 12 linear measurements for the maxillary and mandibular pairs. Figures 4 and 5 show the Bland-Altman plots for the maxillary and mandibular pairs. Both figures graphically demonstrate small systematic errors between the 2 techniques. In the maxilla, the results indicated that the mean difference between the CBCT-generated and the OrthoCAD models was 0.07 6 0.08 mm. In the mandible, the mean difference between the CBCT-generated and the OrthoCAD models was 0.01 6 0.21 mm. Differences within a mean 6 2 SD are small and not clinically important. Therefore, our results suggest that the 2 methods can be used interchangeably. DISCUSSION The use of CBCT technology in the dental office allows for the reproduction of multiple images in all 3 planes of the space and evaluating our patients in 3 dimensions. Acquisition of regular study models from SD Average distance (mm) Low 0.28 0.25 0.34 0.30 High 0.60 0.39 0.61 0.59 Median 0.36 0.32 0.44 0.43 SD DICOM files will, however, greatly enhance the use of CBCT images in daily practice and reduce the need for space and the relative costs. Furthermore, CBCTgenerated models do not depend on impressions and can be reproduced at any time by using the baseline DI- COM images. This method might, therefore, offer a valid alternative to all other commercially available digital model systems. Recent studies have indicated that reliable measurements for tooth size, arch length, space analysis, overjet, overbite, and the Bolton ratio can be obtained with digital models. 4,11,13-19 It was also recently shown that CBCT-generated digital models were as accurate as digital models. 13,14 However, CBCT provides a valid alternative to other digital model systems as long as there is also reliable surface matching. The surface analyses of the OrthoCAD and CBCTgenerated digital models in this study demonstrated that the average linear differences between their surface areas varied between 0.28 and 0.61 mm. Although this might be considered promising, there is clearly a need for improvement in generating digital models from CBCT image scans. Our tolerance-level averages showed that a 90% overlap correlation did not appear until a 1.0 to 1.25-mm alteration on the surface area was reached. A tolerance level of 1.25 mm difference can be problematic, considering the precision of clinical orthodontic applications such as indirect bonding. However, correlation averages are derived from the whole surface area. Ortho- CAD and CBCT-generated models differ from each other significantly, particularly in the gingival areas. Although care was taken when the stereolithography files from the CBCT scans were cropped to leave as much tooth structure as possible above the alveolar bone crest, the remaining differences might have negatively affected the overlap correlation averages within the limitations of our study. These numbers might increase if the June 2012 Vol 141 Issue 6 American Journal of Orthodontics and Dentofacial Orthopedics

Lightheart et al 691 Table III. Overlap amounts for maxillary and mandibular casts at the various tolerance levels Maxillary casts Mandibular casts Low High Average Low High Average 0.25 mm 34.29% 60.49% 45.76% 36.02% 60.01% 42.51% 0.50 mm 64.51% 89.93% 72.83% 65.75% 87.42% 68.82% 0.75 mm 83.28% 98.49% 85.49% 79.59% 95.96% 80.57% 1.00 mm 92.24% 99.69% 92.07% 85.11% 98.41% 86.49% 1.25 mm 95.31% 99.99% 96.72% 89.75% 99.49% 91.19% 1.50 mm 98.47% 99.99% 99.59% 93.72% 99.95% 94.85% Table IV. Means and standard deviations of the linear measurements on the paired OrthoCad and CBCT-generated models with single-measure intraclass correlation coefficients (ICC) Maxillary pairs Mandibular pairs OrthoCad models CBCT-generated models OrthoCad models CBCT-generated models Variable (mm) Mean SD Mean SD ICC Variable (mm) Mean SD Mean SD ICC UR4-UL4 41.33 3.00 41.27 3.01 0.999 LR4-LL4 34.52 3.35 34.50 3.36 0.985 UR5-UL5 46.44 2.93 46.30 2.99 0.997 LR5-LL5 39.38 3.95 39.35 3.88 0.989 UR4-ML 26.24 2.71 26.18 2.72 0.999 LR4-ML 20.59 2.45 20.39 2.25 0.874 UR5-ML 32.58 2.56 32.55 2.53 0.998 LR5-ML 26.30 2.72 26.34 2.70 0.998 UL4-ML 26.64 2.57 26.6 2.55 0.996 LL4-ML 20.82 1.61 20.88 1.59 0.994 UL5-ML 32.89 2.60 32.81 2.58 0.996 LL5-ML 27.05 2.05 27.11 2.03 0.995 gingiva was taken out of the evaluation so that it would be purely a tooth-to-tooth comparison. This might explain why the linear measurements, which are related more to tooth structure, showed a stronger agreement between the 2 techniques. Therefore, tasks that are related to gingival surface areas eg, design of a vacuum-formed appliance should not be performed on CBCT-generated models. Previous research studies comparing agreement between linear measurements made with OrthoCAD-derived models and conventional models have demonstrated acceptable reproducibility and agreement. 11,12,15,16 However, it has also been reported that complicated measurements eg, available space in the mandible, 4 the irregularity index, 17 and maxillary arch length 18 might have significant mean differences between the classic and digital models. All real-world measurements have some degree of error. Possible causes of statistical differences between impression-derived casts and digital models include handling techniques and problems related to different impression materials. 20 Digital models might be adequate but not yet a panacea. 21 In our study, we aimed to document the differences between OrthoCAD and CBCT-generated models for both linear measurements and surface areas. The rationale behind this was to investigate whether CBCT-generated models could be used instead of OrthoCAD digital models. Our results clearly indicate that OrthoCAD and CBCT-generated models have similar values for diagnostic purposes and linear measurements. However, concerns persist in relation to assessment of surface areas; this requires further investigation. CBCT-based diagnostic tasks depend on the accuracy of dental measurements. Baumgaertel et al 22 examined the reliability of dental measurements made on CBCT reconstructions using dry human skulls. Their results indicated that dental measurements from CBCT volumes could safely be used for quantitative analysis. Although they found a slight underestimation of the actual value, this became statistically significant only when several measurements were combined. Similarly, dental measurements of CBCT volumes showed good reliability in determining the diameter of unerupted teeth 23 and tooth and root lengths. 24 Furthermore, Damstra et al 25 concluded that measurements on 3-dimensional surface models of 0.25 to 0.40 voxel size data sets are accurate compared with direct caliper measurements. The Galileos Comfort x-ray imaging unit used in this study has a resolution between 0.15 and 0.30, which is acceptable according to the suggested range. However, a decrease in voxel size might enhance the quality of the scans and result in better correlation of CBCT-generated models with the classic and digital models. Therefore, American Journal of Orthodontics and Dentofacial Orthopedics June 2012 Vol 141 Issue 6

692 Lightheart et al Fig 4. Bland-Altman plot for maxillary linear measurements. Fig 5. Bland-Altman plot for mandibular linear measurements. accuracy of CBCT-generated models obtained by other commercially available scanners with smaller voxel sizes might be worthwhile to investigate in the future. One practical limitation with the use of CBCT images to generate dental models might relate to the radiation dose absorbed by the patient, which we hope can be reduced to a negligible amount in the next generation of cone-beam scanners. Until then, digital plaster casts generated by CBCT imaging should not be obtained from every patient. CONCLUSIONS Within the limitations of this study, overlap amounts between the surface areas of the OrthoCAD and the June 2012 Vol 141 Issue 6 American Journal of Orthodontics and Dentofacial Orthopedics

Lightheart et al 693 CBCT-generated digital models were fair. However, the linear measurements obtained from the CBCT images indicate a high level of accuracy when compared with the OrthoCAD models. Therefore, CBCT-generated digital models can be considered adequate in the early planning phase of orthodontic treatment. REFERENCES 1. Han UK, Vig KW, Weintraub JA, Vig PS, Kowalski CJ. Consistency of orthodontic treatment decisions relative to diagnostic records. Am J Orthod Dentofacial Orthop 1991;100:212-9. 2. Garino F, Garino B. From digital casts to digital occlusal set-up: an enhanced diagnostic tool. World J Orthod 2003;4:162-6. 3. Garino B, Garino F. Comparison of dental arch measurements between stone and digital casts. World J Orthod 2002;3: 250-4. 4. Quimby ML, Vig KW, Rashid RG, Firestone AR. The accuracy and reliability of measurements made on computer-based digital models. Angle Orthod 2004;74:298-303. 5. Rheude B, Sadowsky PL, Ferriera A, Jacobson A. An evaluation of the use of digital study models in orthodontic diagnosis and treatment planning. Angle Orthod 2005;75:300-4. 6. Bell A, Ayoub AF, Siebert P. Assessment of the accuracy of a threedimensional imaging system for archiving dental study models. J Orthod 2003;30:219-23. 7. Joffe L. OrthoCAD: digital models for a digital era. J Orthod 2004; 31:344-7. 8. Mayers M, Firestone AR, Rashid R, Vig KWL. Comparison of peer assessment rating (PAR) index scores of plaster and computer-based digital models. Am J Orthod Dentofacial Orthop 2005;128:431-4. 9. Mah J. The evolution of digital study models. J Clin Orthod 2007; 9:557-61. 10. Kau CH, Richmond S, Palomo JM, Hans MG. Three-dimensional cone beam computerized tomography in orthodontics. J Orthod 2005;32:282-93. 11. Leifert MF, Leifert MM, Efstratiadis SS, Cangialosi TJ. Comparison of space analysis evaluations with digital models and plaster dental casts. Am J Orthod Dentofacial Orthop 2009;136:16.e1-4; discussion, 16. 12. Fleming PS, Marinho V, Johal A. Orthodontic measurements on digital study models compared with plaster models: a systematic review. Orthod Craniofac Res 2011;14:1-16. 13. Kau CH, Littlefield J, Rainy N, Nguyen JT, Creed B. Evaluation of CBCT digital models and traditional models using the Little s index. Angle Orthod 2010;80:435-9. 14. Creed B, Kau CH, English J, Xia JJ, Lee RP. A comparison of the accuracy of linear measurements obtained from cone beam computerized tomography images and digital models. Semin Orthod 2011;17:49-56. 15. Santoro M, Galkin S, Teredesai M, Nicolay O, Cangialosi TJ. A comparison of measurements made on digital and plaster models. Am J Orthod Dentofacial Orthop 2003;124:101-5. 16. Goonewardene RW, Goonewardene MS, Razza JM, Murray K. Accuracy and validity of space analysis and irregularity index measurements using digital models. Aust Orthod J 2008;24:83-90. 17. Stevens DR, Flores-Mir C, Nebbe B, Raboud DW, Heo G, Major PW. Validity, reliability, and reproducibility of plaster vs digital study models: comparison of peer assessment rating and Bolton analysis and their constituent measurements. Am J Orthod Dentofacial Orthop 2006;129:794-803. 18. Mullen SR, Martin CA, Ngan P, Gladwin M. Accuracy of space analysis with emodels and plaster models. Am J Orthod Dentofacial Orthop 2007;132:346-52. 19. Horton HM, Miller JR, Gaillard PR, Larson BE. Technique comparison for efficient orthodontic tooth measurements using digital models. Angle Orthod 2010;80:254-61. 20. Torassian G, Kau CH, English JD, Powers J, Bussa HI, Marie Salas-Lopez A, et al. Digital models vs plaster models using alginate and alginate substitute materials. Angle Orthod 2010;80: 474-81. 21. Brusco N, Andreetto M, Lucchese L, Carmignato S, Cortelazzo GM. Metrological validation for 3D modeling of dental plaster casts. Med Eng Phys 2007;29:954-66. 22. Baumgaertel S, Palomo JM, Palomo L, Hans MG. Reliability and accuracy of cone-beam computed tomography dental measurements. Am J Orthod Dentofacial Orthop 2009;136:19-28. 23. Nguyen E, Boychuk D, Orellana M. Accuracy of cone-beam computed tomography in predicting the diameter of unerupted teeth. Am J Orthod Dentofacial Orthop 2011;140:e59-66. 24. Sherrard JF, Rossouw PE, Benson BW, Carrillo R, Buschang PH. Accuracy and reliability of tooth and root lengths measured on cone-beam computed tomographs. Am J Orthod Dentofacial Orthop 2010;137(4 Suppl):S100-8. 25. Damstra J, Fourie Z, Huddleston Slater JJ, Ren Y. Accuracy of linear measurements from cone-beam computed tomography-derived surface models of different voxel sizes. Am J Orthod Dentofacial Orthop 2010;137:16.e1-6. American Journal of Orthodontics and Dentofacial Orthopedics June 2012 Vol 141 Issue 6