1. Introduction
Intraoral impressions are an important part of everyday work in dental medicine. They serve diagnostic, forensic and treatment purposes. Intraoral impressions can be subdivided into conventional physical impressions and into intraoral digital impressions. Conventional and digital impression systems are used to produce physical study models, physical working models and digital models that can be used for diagnostic purposes such as virtual articulation or direct appliance production [
1,
2,
3].
Intraoral digitization, also called computerized optical impression making (COIM), seems to have become the new method of choice because it provides numerous advantages such as real-time visualization, easier repeatability, no need for cleaning of conventional impression trays, no cast pouring and no wear of the study or working model [
4]. It reduces the number of steps needed to obtain a master cast compared to conventional methods. This direct way of impression making also offers benefits in time efficiency compared to conventional impression methods needed for laboratory scanning processes as well as improved patient comfort [
5,
6,
7]. Apart from the fields of prosthodontics and implantology, computerized optical impression making is increasingly used in orthodontic treatments [
8], for example, for the customization and placement of brackets and for the production of temporary anchorage devices (TADs), clear aligners and retainers [
9,
10,
11,
12,
13,
14,
15,
16,
17].
Nonetheless, to establish a fully digital workflow in clinical practice is challenging. It can include high costs for an intraoral scanner, associated hardware and software. COIM in daily practice usually requires an internet connection, which is not available in every orthodontic clinic. Furthermore, the use of intraoral scanners requires training, which can be particularly difficult for clinicians unfamiliar with operating digital technology [
18]. Moreover, digitally measured tooth crowding can differ from conventionally measured crowding [
19]. Intraoral scanners can provide different levels of accuracy, time efficiency as well as a variety of additional capabilities like automated caries detection. These features depend on the manufacturer, model generation and software version of the intraoral scanner [
6,
20,
21,
22], which leads to different indication spectrums of intraoral scanners. This further increases the complexity of integrating COIM into daily practice. In addition, the current literature shows that full arch scans are still challenging for intraoral scanners, with significantly decreased accuracy compared to partial arch scans [
23].
As the first step along a digital clinical workflow, it is important to use an intraoral scanner which generates optical impressions with high accuracy because digital impressions with a low accuracy impair the accuracy of the digital model used for diagnostic purposes or for appliance design. An impression system was considered to yield the highest accuracy if it showed the lowest trueness and precision values and thus the highest trueness and precision compared to the other intraoral scanners tested in this study. The aim of the trueness evaluations was to determine 3D deviations between the reference datasets which were generated with the aid of an industrial scanner and the oral scanners’ datasets. The term trueness is defined as the closeness of agreement between the average value obtained from a large series of test results and an accepted reference value. It relates to how close the measured size of the scanned teeth is to the real size of the teeth. The real size of the scanned teeth was anticipated with the aid of the industrial scanner. The aim of the precision comparisons was to determine the repeatability of the five intraoral scanners tested in this study. The term precision tells us to what extent the different measurement values of the same measurement system and the same object differ from each other [
24].
4. Discussion
The in vitro approach of this study made it possible to provide similar, controllable scanning conditions for all scanners. In vivo scans are more difficult to perform because of limited space in the oral cavity, patient movements, moisture and soft tissue obstacles. In addition, the oral cavity is too small for the reference scanner used in this study. As reported in another study, extraoral long span scans with intraoral scanners can be more accurate than intraoral long span scans [
25]. This has to be considered when interpreting the results of this study. In order not to influence the intraoral scanners’ accuracy in a negative way, a non-translucent plastic material with natural tooth-colored teeth was chosen for the reference models in this study. More translucent materials such as enamel, which are present during patient care, seem to have a negative effect on the accuracy of many intraoral scanners [
26]. For these reasons, the tested intraoral scanners are expected to provide lower accuracy under clinical conditions.
The reference scanner has been reported to be true to 3 μm [
26,
27,
28], which is important because all intraoral scans have been compared to the reference scans during the trueness evaluations. The scans in this study were performed by a single, experienced dentist in order to not influence the scanning process in a negative way. The method of best-fit 3D superimposition was chosen as the most suitable because this study examined the ability of intraoral scanners to scan natural tooth shapes [
29,
30].
The limit of clinical acceptance for trueness values was set to 50 μm. This limit was chosen because intraoral appliances with higher differences to a patient’s tooth positions can impair blood flow in the periodontal ligament [
31,
32]. This can be seen as the upper limit of clinical acceptance assuming the theoretical fact that no further inaccuracies are added on top of the intraoral scan’s inaccuracy along the clinical workflow.
A major reason for the lower trueness of the CS compared to the TR seemed to be the lower number of datapoints, which can lead to higher calculated deviations of the reference scan file after 3D superimposition and thus to higher trueness values. However, there was no statistically significant difference between the full arch mean trueness of the CS, IT and TR (
p < 0.05,
Table 2,
Figure 5). By analogy with this study, another in vitro study which evaluated the successors of the CS and TR by scanning a full dentate maxilla with unprepared teeth found comparable mean trueness values of 26.9 ± 15.9 μm with the CS 3600, 27.7 ± 6.8 μm with the TRIOS 3 and 20.8 ± 6.2 μm with the TRIOS 4 [
33]. In line with this, an in vivo study evaluating the CS 3600 and the TRIOS 3 in scanning the maxillary arch found comparable trueness levels between the two successor models [
34]. Another in vitro study, however, which evaluated the effect of artificial saliva on COIM of full arch scans with unprepared teeth, found higher trueness values for the CS 3600 and TRIOS 3 than the present investigation [
35]. These findings indicate that proper moisture control is more important than further developed scanner generations.
The PS scanner works with the principle of optical triangulation and thus faced similar technical challenges to the CS. However, the CS can be stabilized on neighboring structures before a single scan image is taken, which has probably contributed to lower trueness and precision values of the CS compared to the PS. The PS constantly has to be moved above the scanned object at a certain focal distance because it automatically takes scan images at a higher rate compared to the CS [
36].
Another examination reported the highest PS distortions in the apicocoronal axis [
37]. In the present study, PS deviations were more evenly distributed over the surface of the teeth. In accordance with the present investigation, other authors also reported surface irregularities and a high number of datapoints for the PS [
30]. The PS showed the highest numbers of outliers in this study (
Table 6). This contributed to higher trueness and precision values of the PS in comparison to the other intraoral scanners tested in this study. The main reason for the lower trueness and precision of the PS compared to the other scanners seemed to be alignment mistakes because the PS showed significantly higher accuracy in the front, left posterior and right posterior segment than in the whole dental arch scans. In a recent study, median 3D full arch trueness values of the latest Planmeca scanner generation were higher than the respective values of the latest generation of TRIOS scanners [
21]. Thus, in accordance with the present study, it can be stated that TRIOS scanners benefited more from generation changes than Planmeca scanners. The PS always showed the lowest accuracy in the front segment, presumably because the front has less geometrical information than the side sectors of the dental arch. This probably contributed to alignment mistakes (
Figure 14). The reason for the better acquisition of maxillary interproximal areas has yet to be identified. More difficult handling, non-uniform tessellation and the highest number of reported outliers in this study have contributed to the lower accuracy of the PS compared to the CS despite the fact that both scanners used the same basic measurement principle of optical triangulation. The PS did not reach a trueness level of clinical acceptance in full arch scans in this study.
Similar to the point-and-click image acquisition mode of the CS, the IT used single images for the data acquisition. In triangulation-based scanners, light reflections from three different angles are measured. Depending on the reflective characteristics of the surface, the reflecting angle and thus the measured distance can be different. In contrast, the IT used the measurement principle of parallel confocal microscopy and only measured light rays that were reflected to its respective focal plane and focused on the optical sensor [
38]. However, the presence of a color wheel can lead to mechanical irritations during the scanning process due to vibrations and to a limited speed of data acquisition due to a limited switching speed of the color wheel [
5]. Furthermore, confocal sensors are sensitive to temperature changes and are only calibrated for a certain temperature. In addition, only certain wavelengths can be measured, which limits the number of possible measurement distances [
39]. Despite these technical challenges, the IT performed among the top three scanners concerning mean accuracy of the full dental arch. In contrast to the PS and TD, the IT showed comparably small differences between full arch and partial arch precision. A certain deviation pattern for iTero with larger deviations at one distal end of the dental arch was reported by other authors [
40]. This trueness deviation pattern could not be confirmed in the study at hand as most higher trueness deviations were located on both ends of the dental arch and in the front area. Another examination reported a resolution of 34.2 datapoints per mm
2 for the iTero, which is very similar to the average resolution determined in this study of 31.3 datapoints per mm
2. By analogy with the present study, no relationship between trueness and resolution was found for the IT [
28]. Despite a higher number of outliers and non-uniform tessellation compared to the CS, the IT showed the lowest maxillary full arch mean trueness values of all intraoral scanners tested in this study. In contrast, the IT showed 44% more blocked interproximal areas than the other tested point-and-click system (CS). Possibly this was due to hardware differences as the CS used active triangulation technology.
In a recent study, the mean mandibular and maxillary trueness values ranged from 32 to 48 μm with the iTero Element 5D, which is the latest iTero model generation [
41]. The higher mean trueness values of the latest iTero model generation compared to the values of the iTero scanner used in this study were probably caused by the circumstance that natural teeth with filling materials were more difficult to scan. Nonetheless, this highlights that newer model generations do not necessarily produce scans with a higher trueness.
An important feature of the TR is that, while scanning, the focal plane is changed in a periodic pattern without changing the scanner’s head distance from the scanned object [
38]. A high contrast is achieved as every point of the scanned surface is only scanned when the point is in focus and when it shows a maximum level of correlation in the scanner’s optical field on the digital sensor. Scanning errors might occur when changing of the focal plane overlaps with changing positions of the operator’s arm. This error is supposed to be minimized by a high speed of focal plane changes and very short data acquisition time [
42]. The TR showed the highest number of datapoints in maxillary scans in the study at hand and the second highest number of datapoints in mandibular scans. However, the TR also exhibited the second highest number of outliers. Despite a higher number of outliers compared to the IT, CS and TD and non-uniform tessellation, The TR showed the highest overall trueness and precision of all intraoral scanners tested in this study. Nonetheless, there was no statistically significant difference between the overall mean trueness and precision of the TR, IT and CS (
p < 0.05). The TR showed significantly less blocked interproximal areas than the other scanner, which used parallel confocal microscopy in this study (IT). A major reason for this might be that the TR is a video-based system, while the IT works with single pictures. Mean trueness values of the TRIOS 3 and 4 in another in vitro study, which also made use of a full arch reference model with unprepared teeth, were higher than with the TRIOS scanner tested in this study [
33]. The TRIOS 4 also produced higher mean trueness values in a recent study [
41]. Thus, contrary to a study by Schmalzl et al. [
21], which used a full arch model with missing and prepared teeth, TRIOS scanners did not show a significant trueness improvement in newer model generations.
Despite the need for reflective coating and the risk of uneven powder application, the TD has previously been shown to work very accurately [
36,
43,
44]. The TD delivers 3D images in a real-time video sequence by using a blue-LED-light wavefront sampling technique. The image acquisition process required a uniform surface, which was provided by reflective coating with titanium oxide powder. Identical reflecting properties were necessary because light reflections from different angles, like in triangulation-based optical sensors, are focused on optical sensors [
36,
45]. The scanner used a rotating device, which is located in the front of the sensor. This device allows images of the same location from different angles and thus works like different cameras taking 2D images of the same object which are computed together to a single 3D picture [
46]. Moving mechanical parts like the built-in rotating aperture and incorrectly applied reflective coating were potential sources of error for the scanning process. On the other hand, reflective coating can also increase scanning accuracy because it creates a more uniform reflecting surface than tooth surfaces composed of different materials [
29]. The missing interproximal parts of TD scans might be due to insufficient powder application. The reference scanner, however, also needed reflective coating and accurately scanned interproximal areas. Thus, insufficient powder application most likely was not the reason for the insufficient acquisition of interproximal areas with the TD. The better acquisition of interproximal areas with the reference scanner might be linked to the different scanning principles of the two scanners. In conclusion, in vitro, apart from interproximal areas, the TD exhibited clinically acceptable trueness only in the front, the left posterior and the right posterior segment. Full arch trueness did not meet the required level of clinical acceptance of 50 μm. The main reason for the TD’s higher whole arch trueness values compared to the individual sector values seemed to be alignment errors.
In general, a higher amount of geometrical information has been reported to lead to intraoral scans with higher accuracy [
47]. On the other hand, digital impressions with large spans are associated with a higher number of stitching errors [
48]. Mandibular trueness in this study in comparison to maxillary trueness was higher for all intraoral scanners except for the PS, which did not show statistically significant differences between maxillary and mandibular trueness (
p < 0.05,
Table 4,
Figure 6). Only the mandibular precision of the IT was lower compared to the scanner’s maxillary precision (
p < 0.05,
Table 4,
Figure 7). The maxillary dental arch is longer than the mandibular arch, and it provides a higher amount of geometrical information because of larger maxillary front teeth. Thus, the number of alignment mistakes due to the longer maxillary dental arch must have outweighed the benefits of a higher amount of maxillary geometrical information in the front region.
The leading cause for the lower full arch trueness of the PS and TD compared to the other intraoral scanners tested in this study seemed to be alignment errors because their individual segment trueness values were significantly lower than their full arch trueness values (
Table 2,
Table 4).
It was not possible to fully explain the different results of the intraoral scanners used in this study because detailed information about the scanners’ hardware and software is not publicly available.