A three-dimensional evaluation of a laser scanner and a touch-probe scanner Anna Persson, BSc,a Matts Andersson, DDS, PhD,b Agneta Oden, MSc, PhD, DDS,c and Gunilla Sandborgh-Englund, DDS, PhDd Karolinska Institutet, Institute of Odontology, Huddinge, Sweden; Nobel Biocare AB, Gothenburg, Sweden Statement of problem. The fit of a dental restoration depends on quality throughout the entire manufacturing process. There is difficulty in assessing the surface topography of an object with a complex form, such as teeth, since there is no exact reference form. Purpose. The purpose of this study was to determine the repeatability and relative accuracy of 2 dental surface digitization devices. A computer-aided design (CAD) technique was used for evaluation to calculate and present the deviations 3-dimensionally. Material and methods. Ten dies of teeth prepared for complete crowns were fabricated in presintered yttriastabilized tetragonal zirconia (Y-TZP). The surfaces were digitized 3 times each with an optical or mechanical digitizer. The number of points in the point clouds from each reading were calculated and used as the CAD reference model (CRM). Alignments were performed by registration software that works by minimizing a distance criterion. In color-difference maps, the distribution of the discrepancies between the surfaces in the CRM and the 3-dimensional surface models was identified and located. Results. The repeatability of both scanners was within 10 mm, based on SD and absolute mean values. The qualitative evaluation resulted in an even distribution of the deviations in the optical digitizer, whereas the dominating part of the surfaces in the mechanical digitizer showed no deviations. The relative accuracy of the 2 surface digitization devices was within 6 6 mm, based on median values. Conclusion. The repeatability of the optical digitizer was comparable with the mechanical digitization device, and the relative accuracy was similar. (J Prosthet Dent 2006;95:194-200.)
CLINICAL IMPLICATIONS The evaluation of the 2 surface digitization devices indicated that the repeatability is comparable and the accuracy is sufficient to serve as input in a manufacturing system for fixed dental prostheses.
n dentistry, all-ceramic restorations are becoming a natural choice in all positions in the dental arch. The introduction of modern technology to manufacture dental restorations has generated opportunities to introduce materials that cannot be manipulated by traditional techniques.1 Francois Duret2 first described computer-assisted production of dental restorations in 1971. During the past decades, the development in the area of computer-aided design/computer-aided manufacturing (CAD/CAM) systems has accelerated. However, the number of reports related to accuracy and precision of CAD/CAM systems remains limited.3-8
Presented in part at the 83rd General Session of the International Association for Dental Research, March 2005 (Arthur R. Frechette Research Award finalist). Supported by Nobel Biocare AB and grants from Karolinska Institutet. a PhD student, Karolinska Institutet, Institute of Odontology. b Chief Scientist, Customized Prosthetics R & D, Nobel Biocare AB. c Associate Professor, Karolinska Institutet, Institute of Odontology; Quality Manager, Nobel Biocare AB, Procera, Stockholm, Sweden. d Associate Professor, Karolinska Institutet, Institute of Odontology.
194 THE JOURNAL OF PROSTHETIC DENTISTRY
The fit of a dental restoration depends on quality throughout the entire manufacturing process. Several factors affect the quality, such as preparation design, surface roughness, impression technique,9 production of a dental cast and, finally, when the restoration is complete, the cementation.10,11 In the process based on CAD/ CAM technology, the chain of transferring geometrical data starts with surface digitization of the preparation.12 However, it is difficult to assess the surface topography of an object with a complex form, such as a tooth, due to the various irregularities and geometric configurations that are unique for each tooth, and the fact that there is no exact reference form. By using a computeraided technique for evaluation of the digitizing equipment, the deviations can be calculated and presented 3-dimensionally.13-16 There are various methods available for digitizing the geometry of a body into a digital form. One of the first CAD/CAM systems launched with dental application was the Procera system (Nobel Biocare AB, Gothenburg, Sweden) based on touch-probe scanning.17 Another method of digitizing is by using optical VOLUME 95 NUMBER 3
PERSSON ET AL
THE JOURNAL OF PROSTHETIC DENTISTRY
Fig. 1. A, Five anterior dies made in presintered zirconia. B, Five posterior dies made in presintered zirconia.
methods based on either laser or white light. The laser or white light is projected onto the object, and the reflected patterns are registered by a digital camera. After the reflections have been tracked in the camera image, 3-dimensional (3-D) points can be obtained using triangulation technology. In general, the advantage of optical methods is the use of a noncontact system that allows the scanning of soft and brittle materials. The optical properties of the object may, however, affect the accuracy of the scan data. The purpose of this study was to evaluate the accuracy and precision of 2 dental surface digitization devices: an optical scanner and a mechanical scanner. Ten different dies of prepared teeth were digitized in both systems. The null hypothesis for this investigation was that digitizing of the dies with the 2 dental surface digitization devices would result in 3-D models of equal quality.
MATERIAL AND METHODS The experimental laser scanner (3Shape A/S, Copenhagen, Denmark) evaluated was a laser line scanner. The scanner consists of a table with a model holder and a laser and has high-resolution digital cameras that acquire images of the line as it is projected onto the object. The model to be digitized is fixed in the holder. To ensure complete coverage of the object’s geometry, the table can be rotated and tilted and can move along a horizontal axis. The image-processing software (ScanIt, MARCH 2006
version 4.0.1; 3Shape A/S) processes the images and calculates, by triangulation, a point cloud as a 3-D model. The surface creation software automatically optimizes the data and reduces the number of points in the point cloud. By combining the points to a 3-D polygonal model, a 3-D surface model is automatically created. A commercially available touch-probe scanner, (M50; Nobel Biocare AB) was used. The accuracy of 610 mm was confirmed by Persson et al.4 The die to be digitized was oriented vertically in the holder of the scanner. A sapphire ball forms the tip of the scanner probe that contacts the surface of the die as it rotates around a vertical axis. The probe was positioned, with light pressure, at a 45-degree angle to the axis of rotation of the die. The scanner was set to collect 1 data point at every degree around the circumference of the die. During each turn, the probe was continuously elevated 200 mm, and the entire surface of the die was registered.17 The resulting point cloud was processed with the software provided (Procera System C3D, version 1.4; Nobel Biocare AB). The 3-D model was calculated using an offset of the point cloud, based on the radius of the sapphire ball. A 3-D surface model was automatically created by combining the points to a 3-D polygonal model. The digitizers were calibrated according to the manufacturers’ recommendation. Ten different dies were selected from files in the Procera production. All of the dies had preparations with either chamfer margins or deep chamfer margins for complete crowns. The dies were divided into 195
THE JOURNAL OF PROSTHETIC DENTISTRY
PERSSON ET AL
Fig. 2. Example of process chain for alignment of 2 different readings. (A) Point cloud in CRM from touch-probe scanner; (B) 3-D surface model from laser scanner; (C) Aligned models presenting distribution of discrepancies (positive values imply that surface of CRM located inside 3-D surface model); (D) 2-D cross-section of aligned models.
2 groups based on their shape. Five dies had the shape based on prepared anterior teeth (Fig. 1, A), and the other 5 were based on posterior teeth (Fig. 1, B). The dies were manufactured by Nobel Biocare AB in presintered yttria-stabilized tetragonal zirconia (Y-TZP). This material was chosen for having good surface hardness as well as optical properties that allowed it to be digitized with both digitizers.
Alignment Three readings of each die were performed in both digitizers. Alignments were performed by software (NSI Registration, version 1.1; 3Shape A/S) that works by minimizing a distance criterion. The number of points in a point cloud from 1 reading was calculated and used as the CAD reference model (CRM) (A in Fig. 2). The CRM was aligned to the 3-D surface model of another reading of the same die (B in Fig. 2). In color-difference maps, the distribution of the discrepancies between the surface in the CRM and the 3-D surface models was identified and analyzed (C and D in Fig. 2). The result was presented as absolute mean, standard deviation, median values, and as the 95th percentile of the deviations. To avoid interference of the points below the margin, these were manually removed from the point clouds that were to be used as the CRM. The placement of the starting point 196
was manually chosen, and the alignments were performed automatically.
Interpretation of repeatability and relative accuracy For evaluation, the shortest distances from the points in the CRM to the surface of the 3-D surface models were measured in the registration software. The influence on the readings by the size and shape of the preparations was evaluated. The repeatability of each scanner was assessed by using all 3 readings as CRM, which was aligned to the other 2 readings from the same scanner. The various combinations of alignments were made to identify if any reading differed from the others. Thus, each die resulted in 6 different alignments for each of the 2 scanners. The SD and absolute mean values of the discrepancies were used for the quantitative analysis. Qualitative evaluation was made by identifying and locating the deviations between the surfaces in colordifference maps. To assess the relative accuracy of the 2 surface digitization devices, the CRM from 1 scanner was aligned with the 3-D surface models from the other scanner and vice versa. Each die resulted in 6 different alignments per scanner. Due to skewed distributions, the median values and the 95th percentile of the discrepancies were used for the quantitative analysis. For qualitative VOLUME 95 NUMBER 3
PERSSON ET AL
THE JOURNAL OF PROSTHETIC DENTISTRY
Fig. 3. Repeatability: SD of discrepancies between 3 CRMs from each die aligned with other 2 readings.
Fig. 4. Repeatability: absolute mean of the discrepancies between 3 CRMs from each die aligned with other 2 readings.
evaluation, color-difference maps were used to analyze whether the 2 scanners differed in readability of different shapes, and if so, how.
scanner resulted in smaller deviations than those from the laser scanner. The repeatability of the laser scanner presented as the SD of the distance between the CRM and the 3-D surface model resulted in deviations between 6.3 and 9.1 mm. In the touch-probe scanner, the corresponding figures were 1.9 to 7.5 mm (Fig. 3). The repeatability presented as the absolute mean values of the deviations between the digitized and aligned models ranged from 4.9 to 7.2 mm in the laser scanner and 1.4 to 5.8 mm in the touch-probe scanner (Fig. 4). The qualitative evaluation presented in Figure 5 showed that alignments of dies from the laser scanner resulted in both positive and negative deviations, evenly distributed over the entire surface of the die. Alignments of models from the touch-probe scanner resulted in the dominating part of the surfaces showing no deviations. However, in the concave surfaces there were negative discrepancies. In the anterior group these were located in the proximal surfaces, whereas in the posterior group there was an almost circumferential distribution of the deviations. Figure 5 (B) shows 1 of the alignments performed on dies from the touch-probe scanner, which resulted in the smallest deviation. However, the largest discrepancies were obtained in alignments performed on registrations from the laser scanner, also from the anterior group (C in Fig. 5).
RESULTS The number of points in the CRM differed considerably. In the laser scanner, the dies from the anterior region had approximately 5000 points, whereas the dies from the posterior region had approximately 6500. The corresponding figures for the touch-probe scanner were 23,000 in the anterior region and 19,500 points in the posterior region.
Repeatability The repeatability of both scanners was within 10 mm, based on SD and absolute mean values. As shown in Figure 3, the distribution of the discrepancies varied between the 2 surface digitization devices. In the laser scanner, they were closely assembled but showed a larger numerical value, whereas in the touch-probe scanner the discrepancies were smaller, but the distribution was more scattered. In regard to the 2 groups of dies from anterior or posterior regions, both scanners showed dissimilarities in the distribution of the discrepancies. Overall, alignments made on dies from the anterior region digitized in the laser scanner resulted in the largest deviations, whereas the smallest discrepancies were obtained in alignments from the anterior dies digitized in the touch-probe scanner. For the posterior group, the alignments performed on readings by the touch-probe MARCH 2006
Relative accuracy The relative accuracy of the 2 surface digitization devices was within 66 mm, based on median values of 197
THE JOURNAL OF PROSTHETIC DENTISTRY
PERSSON ET AL
Fig. 5. Repeatability: examples of color-difference maps. (A) and (B) represent small deviations; (C) and (D) represent larger deviations.
2.1 mm for the posterior dies (Fig. 6). The 95th percentile of the deviations ranged from 242 to 29 mm. When using the CRM from the touch-probe scanner, the corresponding figures were 23.7 to 0.6 mm for the anterior and 21.0 to 5.0 mm for the posterior models (Fig. 6). The 95th percentile of the deviations was within the range of 233 to 51 mm. In the qualitative evaluation, when using the dies from the laser scanner as CRM, concave surfaces generally resulted in negative deviations, whereas convex surfaces resulted in positive deviations (A and C in Fig. 7). A mix of both negative and positive deviations was obtained on the surfaces of the axial walls. When using the models from the touch-probe scanner as CRM, negative deviations were generally obtained in convex surfaces, whereas in concave surfaces and in the surface of the axial walls, positive deviations were found (B and D in Fig. 7). However, in the rounded slope of the chamfer, negative deviations were obtained. In the anterior group these were located in the proximal area, whereas in the posterior group the deviations were almost circumferential. Fig. 6. Relative accuracy: median values of discrepancies between 3 CRMs from each die in 1 scanner aligned with two 3-D surface models from other scanner.
the deviations between the CRM and the 3-D surface models (Fig. 6). When the CRMs from the laser scanner were used, the median values of the deviations ranged from 20.9 to 3.4 mm for the anterior and from 25.3 to 198
DISCUSSION The null hypothesis that the 2 surface digitization devices would generate 3-D models of equal quality was accepted. Thus, the laser scanner has the potential to serve as input in a manufacturing system since the touch-probe scanner is currently used in the clinical VOLUME 95 NUMBER 3
PERSSON ET AL
THE JOURNAL OF PROSTHETIC DENTISTRY
Fig. 7. Relative accuracy: examples of color-difference maps presenting characteristic deviations. (A) and (B) represent anterior group; (C) and (D) represent posterior group.
setting and generates satisfactory clinical fit.18 The present study shows that the repeatability of the 2 surface digitization devices is within the range of 10 mm. Based on median values, the relative accuracy of both scanners is within 66 mm. The deviations between the CRM and the 3-D surface models revealed differences in the repeatability of the 2 scanners. The number of points in the CRM differed. Digitizing by the laser scanner resulted in a point cloud with fewer points that had less dense distribution than those digitized by the touch-probe scanner. Independent of the shape of the die, registrations by the laser scanner resulted in an even distribution of the points in the CRM. Thus, the posterior dies resulted in a larger number of points in the CRM compared to the anterior dies. The precision of the laser scanner may be improved by increasing the resolution in certain critical surfaces. In the touch-probe scanner, the shape of the die had an opposite effect on the distribution of the points. In the registrations of the anterior shapes, the resulting point clouds were denser than for the posterior shapes. The same number of points per turn was used and, as a consequence, a larger radius of the die resulted in a less dense point cloud. The helical motion of the probe was influenced by its angle toward the surface. As a consequence, the distribution of the points in the rounded slope of the chamfer was less dense (A in Fig. 2). As a result, this surface showed deviations in the evaluation of the MARCH 2006
repeatability. Previous studies have shown that these portions of the prepared tooth are associated with less than optimal internal fit of the manufactured crown.18 In the present study, the repeatability was correlated to the number of points in the CRM; that is, alignments of anterior dies by the touch-probe scanner resulted in the best repeatability. In the laser scanner, the repeated registrations resulted in very similar alignments. The surface creation software of the laser scanner reduces and optimizes the number of points, whereas in the touchprobe scanner, no such filtration is performed. This could partly explain why the discrepancies are more scattered in alignments from the touch-probe scanner. In this study, the relative accuracy was evaluated since none of the digitizers was a measuring device. To achieve a more accurate CRM to be used as a master model, the number of points should be increased in critical surfaces, so the distance between the points is the same regardless of the shape and radius of the object to be measured. In 2 aligned files, the resulting discrepancy is not only the real error between the files, but also a result of the insecurity of the alignment. The qualitative evaluation indicates that the laser scanner technique tends to round off sharp edges, such as the margin, whereas the touch-probe technique is more efficient in reproducing edged surfaces (D in Fig. 2). The discrepancies that were obtained in the concave surface revealed that the surface from the laser scanner was located outside the surface from 199
THE JOURNAL OF PROSTHETIC DENTISTRY
the touch-probe scanner. Possible reasons for this observation may be that the optical quality of the material in the die was not optimal, or that the resolution of the surface digitization device was not high enough. This will be investigated in future studies. The negative deviations obtained in the proximal surfaces, where the surface from the touch-probe scanner was located outside the surface from the laser scanner, are most likely due to the fact that the elevation of the probe was too steep and, therefore, returned an inadequate distribution of the points in the point cloud by the touch-probe scanner. Thus, the distribution of the deviations from the evaluation of the repeatability seems to have an influence on the outcome of the relative accuracy. The inaccuracies of the 2 scanners have dissimilar characteristics, which affect the deviations in matching. The precision of fit for the prosthetic restoration is dependent not only on the scanner but also on the manufacturing process, including all preceding procedures such as preparation, impression, and fabrication of dental cast. The surface digitization of the prepared tooth is simply 1 step of the entire process. How this input into the CAM system is processed thereafter may affect the quality of the definitive prosthetic restoration, and the following procedures will likely have an impact on the quality of the internal fit. Proposed future studies include the manufacture of crowns, based on the 2 types of scanners, and comparison of internal fit. The methodology used in the present study may be used for this evaluation.
CONCLUSIONS Within the limitations of this study, the following conclusions were drawn: 1. The repeatability and accuracy of the experimental optical digitizer was comparable with the touchprobe surface digitization device. 2. The results showed that a nontouching system has a good potential to serve as input in a manufacturing system for fixed dental prostheses.
PERSSON ET AL
3. Hewlett ER, Orro ME, Clark GT. Accuracy testing of three-dimensional digitizing systems. Dent Mater 1992;8:49-53. 4. Persson M, Andersson M, Bergman B. The accuracy of a high-precision digitizer for CAD/CAM of crowns. J Prosthet Dent 1995;74:223-9. 5. Luthardt RG, Sandkuhl O, Herold V, Walter MH. Accuracy of mechanical digitizing with a CAD/CAM system for fixed restorations. Int J Prosthodont 2001;14:146-51. 6. Dahlmo KI, Andersson M, Gellerstedt M, Karlsson S. On a new method to assess the accuracy of a CAD program. Int J Prosthodont 2001;14: 276-83. 7. DeLong R, Heinzen M, Hodges JS, Ko CC, Douglas WH. Accuracy of a system for creating 3D computer models of dental arches. J Dent Res 2003;82:438-42. 8. Coli P, Karlsson S. Precision of a CAD/CAM technique for the production of zirconium dioxide copings. Int J Prosthodont 2004;17:577-80. 9. Luthardt RG, Koch R, Rudolph H, Walter MH. Qualitative computer aided evaluation of dental impressions in vivo. Dent Mater 2006;22:66-76. 10. Wang CJ, Millstein PL, Nathanson D. Effects of cement, cement space, marginal design, seating aid materials, and seating force on crown cementation. J Prosthet Dent 1992;67:786-90. 11. Wolfart S, Wegner SM, Al-Halabi A, Kern M. Clinical evaluation of marginal fit of a new experimental all-ceramic system before and after cementation. Int J Prosthodont 2003;16:587-92. 12. Luthardt R, Weber A, Rudolph H, Schone C, Quaas S, Walter M. Design and production of dental prosthetic restorations: basic research on dental CAD/CAM technology. Int J Comput Dent 2002;5:165-76. 13. Bloem TJ, Czerniawski B, Luke J, Lang BR. Determination of the accuracy of three die systems. J Prosthet Dent 1991;65:758-62. 14. Brosky ME, Pesun IJ, Lowder PD, Delong R, Hodges JS. Laser digitization of casts to determine the effect of tray selection and cast formation technique on accuracy. J Prosthet Dent 2002;87:204-9. 15. Rudolph H, Quaas S, Luthardt RG. Matching point clouds: limits and possibilities. Int J Comput Dent 2002;5:155-64. 16. Luthardt RG, Kuhmstedt P, Walter MH. A new method for the computer-aided evaluation of three-dimensional changes in gypsum materials. Dent Mater 2003;19:19-24. 17. Andersson M, Razzoog ME, Oden A, Hegenbarth EA, Lang BR. Procera: a new way to achieve an all-ceramic crown. Quintessence Int 1998;29: 285-96. 18. Kokubo Y, Ohkubo C, Tsumita M, Miyashita A, Vult von Steyern P, Fukushima S. Clinical marginal and internal gaps of Procera AllCeram crowns. J Oral Rehabil 2005;32:526-30. Reprint requests to: MS ANNA PERSSON KAROLINSKA INSTITUTET, INSTITUTE OF ODONTOLOGY PO BOX 4064 ALFRED NOBELS ALLE´ 8 HUDDINGE SE-14104 SWEDEN FAX: 46-8-746-7915 E-MAIL: [email protected]
0022-3913/$32.00 Copyright Ó 2006 by The Editorial Council of The Journal of Prosthetic Dentistry.
REFERENCES 1. Andersson M, Oden A. A new all-ceramic crown. A dense-sintered, highpurity alumina coping with porcelain. Acta Odontol Scand 1993;51: 59-64. 2. Duret F, Blouin JL, Duret B. CAD-CAM in dentistry. J Am Dent Assoc 1988;117:715-20.
VOLUME 95 NUMBER 3