Trueness of CAD/CAM digitization with a desktop scanner – an in vitro study

Background Desktop scanners are devices for digitization of conventional impressions or gypsum casts by indirect Computer-Aided Design/Computer-Assisted Manufacturing (CAD/CAM) in dentistry. The purpose of this in vitro study was: 1, to investigate whether virtual models produced by the extraoral scanner have the same trueness as sectioned casts; and 2, to assess if digitization with an extraoral scanner influences the surface information. Methods A polimethyl-methacrilic acid (PMMA) cast and a reference scanner (TwoCam 3D, SCAN technology A/S, Ringsted, Denmark; field of view 200 mm, resolution 0.1 mm ± 0.025 mm) were used to create the reference data in standard tessellation format (STL). According to the extraoral CAD/CAM digitization steps, impressions, mastercasts, and sectioned casts were made, and STL files were generated with the reference scanner. The pivotal point of the study was to digitalize these sectioned casts with the extraoral scanner (Straumann CARES Scan CS2 Visual 8.0 software, InstitutStraumann AG, Basel, Switzerland) and STL files were exported. Virtual caliper measurements were performed. Absolute deviations were compared using multilevel mixed-effects linear regression. Relative distortions were calculated with mean absolute errors and reference values. Results Differences were observed in measurements of tooth sizes. All four prepared teeth were affected. No relationship was observed in relative deviations. Absolute differences between all the indirect digitization steps considering arch distances were: impressions, − 0.004 mm; mastercasts, 0.136 mm; sectioned casts, − 0.028 mm; and extraoral scanner, − 0.089 mm. Prepared dies on the virtual casts (extraoral scanner) were closer to each other than those on the sectioned gypsum casts. Relative deviation calculations revealed no relationship with the position of the dies in the arch. Conclusion The trueness of the virtual models generated by the extraoral scanner system used in this study was different from the dimensions of the sectioned casts. The digitization of gypsum casts changes both the dimensions of dies and the distances between the dies. The virtual casts had smaller distances than any distances measured at previous steps. Either bigger dies or longer distances did not result in greater distortions. We cannot, however, generalize our results to all scanners available on the market, because they might give different results.

(such as, for example, the lost-wax process) [7][8][9][10]. As the first step, a scanner (either extraoral or intraoral) is utilized to transfer the information of the oral cavity into the computer. Extraorally, indirect CAD/CAM digitization is usually performed by scanning gypsum casts, although sometimes conventional impressions may also be scanned [7,11]. The intraoral scanning technique performs a direct scanning in the oral cavity [8,  12]. Some studies assess the patients' perception and treatment comfort, clinical outcomes, and time efficiency [5,13], while others compare the clinical accuracy of conventional impressions with that of direct digitizers [9,14,15] or the accuracy of direct versus indirect digitizers [16][17][18].
The accuracy measurement based on the ISO (International Organization for Standardization) standard 5725 [14,[16][17][18] has two components: precision and trueness. In a previous study, our research group compared the accuracy of three intraoral scanners to a desktop scanner [18]. One of the results was that the desktop scanner was less accurate compared to the intraoral scanners. This raised a question: which step or steps of the indirect CAD/CAM digitization changed the original surface information? The aim of this study, therefore, was to evaluate the trueness of virtual models produced by the extraoral scanner. Our hypothesis was that there is no significant difference between the sectioned casts and the virtual casts made by the extraoral scanner.

Study design
Figures 1a and b depict the study design. A polimethylmethacrilic acid (PMMA) cast and a reference scanner (TwoCam 3D, SCAN technology A/S, Ringsted, Denmark; field of view 200 mm, resolution 0.1 mm ± 0.025 mm) were used to create the reference data in standard tessellation format (STL). According to the extraoral CAD/CAM digitization steps, impressions, mastercasts, and sectioned casts were made, and STL files were generated with the reference scanner. The pivotal point of the study was to digitalize these sectioned casts with the extraoral scanner (Straumann CARES Scan CS2 Visual 8.0 software, Institut Straumann AG, Basel, Switzerland) and STL files were exported. Accordingly, in order to evaluate the information changes from the reference data through the steps of indirect digitization up to the CAD/CAM scan made by the extraoral scanner, we strictly adhered to the steps of indirect CAD/CAM digitization and compared the data of impressions, mastercasts, sectioned casts, and desktop scanner digitization to the reference data.

Reference model
A PMMA cast was made from an upper arch, holding the original information of hand-prepared teeth #14, #21, #24, #27 with shoulder preparation for all-ceramic restorations. Two edentulous areas are represented on the arch, #13-21 and #24-27 (Fig. 2).

Reference scanner and reference virtual model
For the high-precision data acquisition, a point-laser scanner (635 nm wavelength, 1 mW power, Class IEC 2) was utilized (TwoCam 3D, SCAN technology A/S, Ringsted, Denmark). This scanner uses a double triangulation technique with the following parameters: field of  The reference model was scanned by the point-laser scanner, and an STL reference virtual model was generated (Fig. 3).  [10,23,24,[31][32][33] (Fig. 4).

Impression
In accordance with prevailing in vitro temperature conditions, the recommended setting time (5 min and 30 s) was doubled to ensure the correct setting of the material [32,33]. After setting, the impression was removed from the cast and washed for 5 min. Disinfection with Zeta 7 spray (Zhermack, Zhermack Spa, Badia Polesine, Italy) followed. At least 1 h but not more than 24 h after disinfection, each impression was scanned once with the reference scannerthus, a total of 10 impression scans were performed. The produced STL files were saved.

Mastercasts and sectioned casts
Not more than 24 h after taking the impressions, the 10 impressions were casted in the dental laboratory with type IV gypsum (GC Fujirock EP, GC Corp., Tokyo, Japan) [25,28,30]. The mixing of the gypsum was performed with distilled water (100 g/25 ml) first by hand, then with a vacuum mixer (20 s, BEGO Motova SL, BEGO USA Inc., Lincoln, RI, USA). A mechanical vibrator (WASSERMANN Rüttler KV-26, Wassermann Dental-Maschinen GmbH, Hamburg, Germany) was used (6000 rpm, 0.4 mm) to produce the casts. The setting time of the gypsum was always 1 h. After setting, the mastercasts were removed and finalized (Figs. 5 and 6). All 10 mastercasts were digitized with the reference scanner, and the STL files were saved. Next, the mastercasts were sawed in the laboratory (Giroform, Amann Girrbach GmbH, Pforzheim, Germany), and 10 sectioned casts were scanned with reference scanner (Fig. 7).
To perform CAD/CAM scans used in CAD/CAM technology [12], the sectioned casts were scanned 24-72 h after the casting of the impressions with an extraoral scanner (Straumann CARES Scan CS2 Visual 8.0 software,   This involves a first full-arch scan of the gypsum cast followed by a second scanning of the prepared dies. The software aligns the die scans onto the full arch scan, and this alignment will lead to higher precision of the scanned data. The STL files of 10 final virtual casts were exported and saved.

Superimposition and virtual digital caliper
For the virtual comparisons, a best-fit alignment algorithm and virtual caliper tool were used in Geomagic Verify software (3D systems, 333 Three D Systems Circle, RockHill, SC, USA). Eleven virtual caliper measurements were performed on the reference virtual model as follows: a section plane was placed on the virtual casts, and on this plane, mesial-distal points and buccal-oral points of the prepared teeth and 6 points for the measurements of abutment distances were appointed (Fig. 9). Each impression, mastercast, and sectioned cast STL file made by the reference scanner and the STL files from the sectioned casts made by the extraoral scanner were imported and aligned one after the other to the virtual reference model by best-fit alignment. The Geomagic software calculated the differences between the distances measured on the reference versus the superimposed data and output the resulting data in Excel.

Statistics
Observation of differences between the data points was performed in two ways. First, absolute deviations and differences were calculated (in mm). Absolute deviations were compared across CAD/CAM steps using multilevel mixed-effects linear regression and were interpreted as differences in trueness. Explanatory variables included a categorical indicator for each CAD/CAM step and a random intercept term at the observation series level. The model allowed for intragroup correlation between observations within the same observation series. Models were fitted separately for each location. Differences between CAD/CAM steps were expressed as point estimates of the fixed effect, 95% confidence intervals (CI), and p values. Second, relative distortions were calculated. To avoid misrepresentations arising from calculating the averages of positive and negative values (which can result in smaller deviations), absolute values of the observed distances were used. For each die diameter and arch distance, a mean absolute error was calculated. A relative distortion was calculated with mean absolute error and reference value registered with the reference scanner (relative distortion = mean absolute error/reference value). Median and interquartile range values were used because of skewness. Stata was used for data management and analysis (StataCorp. Stata Statistical Software: Release 15. College Station, Texas: StataCorp LLC).

Differences in the measurements of tooth sizes
Significant differences were observed in the measurements of tooth diameters between the steps of indirect CAD/CAM technology performed with the Straumann extraoral scanner at all eight locations of interest (at a minimum of one and a maximum of four per location) according to the absolute deviations ( Fig. 10 and Tables 1  and 2). Overall, the highest number of differences were as follows. Between the sectioned cast and the extraoral scanner, significant differences (either higher or lower) were observed in six of the eight locations (p < 0.01), and the data measured on the impressions were significantly different from the data measured on the extraoral scanner at five locations (p < 0.05). All four prepared teeth were affected by these differences. There were differences between the steps of indirect CAD/CAM, but no relationship was observed in relative deviations ( Fig. 11 and Table 3), meaning that longer diameters did not result in greater distortions.

Differences in the measurements of arch distances
Statistically significant differences were observed between all the indirect CAD/CAM digitization steps considering arch distances ( Table 6). The distance between the dies became larger after the impressions (median − 0.004 mm, IQR = 0.198) were poured with gypsum and the mastercasts (0.136 mm, IQR = 0.157) were made. By sectioning the mastercast, the distances became smaller and the values of the sectioned casts (− 0.028 mm, IQR = 0.279) were similar to those of the impressions. Virtual casts (− 0.089 mm, IQR = 0.322) made by the extraoral scanner showed smaller distances compared to any of the previous steps' values ( Fig. 12 and Table 4).

24-27 inside
Distance measurement on the closest points of 24-27 showed the following: VPS impressions had the trueness of 0.006 mm (IQR = 0.071), while mastercasts had 0.149 (IQR = 0.034) and sectioned casts had − 0.023 mm (IQR = 0.073). Virtual casts made by the extraoral scanner had the trueness of − 0.086 mm (IQR = 0.043).    The analysis revealed significant distortion in trueness at all steps of the indirect CAD/CAM method, and significant differences were found between the impressions and virtual casts considering all three measured distances (p < 0.01, Table 6). The smallest values were registered on the virtual casts made by the extraoral scanner, which were significantly smaller than the values of the impressions and the sectioned casts.

24-27 outside
Overall, absolute deviation calculations showed that the dies on the virtual casts made by the extraoral scanner were larger than those on the sectioned gypsum casts. However, prepared dies were closer to each other on the virtual casts made by the extraoral scanner compared to the sectioned casts made by the reference scanner ( Fig. 13 and Table 5). Relative deviation calculations revealed no relationship with the position of dies in the arch (Fig. 14 and Table 6 and 7), meaning that further distances did not result in higher distortions.

Discussion
Little is known about the scanning step in indirect CAD/CAM digitization, and our analysis is one of the first studies to shed more light on the entire process: not only the impression, mastercast, and sectioned cast, but also the digitization step. Our null hypothesis was rejected: our results show that there are differences between the dimensions represented by the sectioned cast and the virtual cast made by the optical extraoral scanner used in the study. We also showed that the scanning procedure influences the diameter of the dies and that there is no relationship between the distance of the dies and the distortion in the measurement of the arch.
Accurate virtual models are necessary to make accurate prostheses. Usually, indirect CAD/CAM digitization is based on conventional impressiontaking and stone-cast making procedure. Residual stresses are well known as a feature of gypsum crystallization. ADA Specification #25 describes that the expansion of the dental gypsum might be as much as 0.2% [38]. This distortion caused by expansion can be compensated for by sectioning the cast. In this case, the dies must be in the correct position in the arch, and they must be able to be removed [27]. Sectioning can release stress in the gypsum so that the dies can be repositioned to the original position granted by the acrylic base [30]. The distortion of the conventional cast making steps represented in this study are similar to what has been found in other studies [25,28,30]. There have been studies that evaluate the trueness and precision of extraoral scanners by using solo abutments [35], silicone impression material [11], or by an easy-to-measure metric cast [34], but there is hardly any information about the factors that influence accuracy between the making of the gypsum cast and the CAD/CAM production of  Our study aimed to assess the trueness of indirect digitization of gypsum casts using an optical extraoral scanner, whereby we hypothesized that there was no difference between the sectioned casts and the virtual casts made by the extraoral scanner. However, when we explored the differences between the trueness of sectioned gypsum cast and the virtual cast made by the optical extraoral scanner, our null hypothesis was rejected. Vanderweghe et al. [34] described how gypsum casts are harder to digitize with optical scanners because of the rough surface. In their study, three out of four scanners had better trueness at scanning acrylic resin. Based on this, the sensitivity of the scanner might also play an important role during scanning. On the other hand, studies show that there is no correlation between triangle numbers (number of digital points) and accuracy, but  accuracy depends on the quality of the point cloud generated by the software algorithm [35,39]. The diameters of the dies were influenced (distorted) by the extraoral scanner in our study. The differences observed on all four prepared teeth #11, #14, #24, #27 during the extraoral scanner scanning step indicates a necessity of caution during everyday dental practice. Our observed distortions, however, differ from those detected in other studies. Wöstmann et al. [40] compared the accuracy of four intraoral and 10 extraoral digitizers in their study, with the reference model die shaping a chamfer-prepared canine and a chamferprepared molar. They found differences between the accuracy of intraoral and extraoral scanners with an accuracy below 20 μm. Jeon et al. [41] measured accuracy during extraoral scanning of impressions from prepared teeth for all-ceramic restorations. Their results showed the trueness to be less than 30 μm. The reason for this difference in data could be that these studies measured single prepared teeth only, while in our study two-step scanning procedures were performed on gypsum casts to obtain the whole arch and the dies information as well. However, these differences between the results of the present study and those of other studies are not clinically significant.   They used a geometrical model with rectangular and cylindrical shapes as a reference model, and the extraoral scanners were tested with acrylic based casts (mean trueness: 0.047 mm) and gypsum casts (mean trueness: 0.099 mm). Most of their scans met the requirements of clinical accuracy published in other studies [20][21][22], indicating a lack of clinically unacceptable distortion caused by indirect CAD/CAM digitization. Their results are somewhat different from the results of our present study (our mean trueness: − 0.086 mm). These differences may have arisen from the fact that the two studies used different methods: in addition to us using different casts, the reference model of our study was a PMMA replica of hand-prepared anatomical teeth. As such, the   anatomical arch and the prepared teeth in our study may be more difficult for the extraoral scanner to scan that the cylindrical shape in their study. One limitation of our study is that the trueness in our study might be considered low. Mandelli et al. [35] utilized seven extraoral scanners, although not the Straumann CARES Scan CS2 that we used for scanning, and most of their scanners used only a single-step scanning procedure, while we used a two-step scanning procedure. Furthermore, their reference die was an easy-toscan solo non-anatomical titan abutment, as opposed to an anatomical gypsum cast with hand prepared abutments used in our study. Their accuracy varied between 8 and 30 μm, depending on the scanner system while our trueness values were much higher. The two-step scanning procedure combined with a hand-prepared abutment with a more complex surface might explain the lower trueness in our study. Another limitation is that our study used only the digitization of a sectioned gypsum cast, and we did not assess the scanning of a precisional-situational impression using a laboratory extraoral scanner or intraoral scanners [7]. However, the scanning of conventional impressions is performed much less often than the scanning of sectioned gypsum casts, and therefore our study focused on the more commonly used indirect digitization. Last but not least, our study focuses only the Straumann extraoral scanner, therefore these results cannot be generalized for all extraoral desktop scanners. However, according to Holst et al. [34] who compared optical and contact scanner accuracy, there is no significant difference between the two types of scanners. A further limitation is that we measured only trueness but not precisionmeasuring precision might be the topic of future studies. Finally, the laboratory scanner used in this study was not precalibrated specifically for the purposes of this study. However, everyday dental laboratory work uses limited calibrations, and therefore our protocol followed a reallife approach.

Conclusions
Within the limitations of the present in vitro study, we can conclude that the trueness of the virtual models generated by the extraoral scanner system used by us in the study is different compared to the dimensions of the sectioned casts. The digitization of gypsum casts changes both the dimensions of the dies and distances between the dies. The differences observed on all four prepared teeth were both positive and negative at the scanning step. No relationship was observed in relative deviations, meaning that higher values of the dies did not result in higher distortions. At the last step of indirect CAD/ CAM digitization, the distances of the virtual casts made by the extraoral scanner were smaller than any of the distances measured at previous steps. No relationship was revealed with the position of dies in the arch, meaning that further distances did not result in greater distortions. Distortions observed at half arch distance may have clinical relevance. We cannot, however, generalize our results to all scanners available on the market, because they might give different results. Therefore, future studies may further explore the accuracy of other extraoral scanners in life-like samples. Authors' contributions GJ-K contributed to the design of the study, designed the analysis plan, performed data management, generated most of the figures, and wrote the paper. BV contributed to the design of the study, performed data analysis, and contributed to writing the paper. SzK contributed to the design of the analysis, performed data acquisition and data management, and contributed to writing the paper. VAG contributed to the interpretation of the data and the writing of the paper. JB contributed to the design of the study, was involved in supervising data collection, conceived the analysis plan, and contributed to writing the paper. PH contributed to the design of the study, and provided advice, was involved in supervising data collection, and contributed to writing the paper. All authors read and approved the final manuscript.

Funding
The present in vitro study was not funded by any grant.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.