The results of this study indicated that dimensional variations in crowns and inlays significantly affect the accuracy of both indirect and direct digital impression techniques. Consequently, the null hypothesis is rejected.
Statistically, the average crown length of human permanent teeth ranges from 7.1 to 11.1 mm, whereas the average crown width ranges from 5.4 to 11.1 mm. Furthermore, an effective preparation design for crowns and inlays is essential to achieve optimal outcomes in restorative dentistry. This includes facilitating adequate support under stress conditions and ensuring a rational distribution of stresses to enhance fracture resistance21,22. The actual sizes of the prepared crowns and inlays can vary owing to differences in the dental conditions among clinical patients. Therefore, in this study, the model parameters were adjusted by ± 2 mm in accordance with ADA 132 to accurately simulate the various sizes of crowns and inlays that may be encountered in a clinical setting.
According to Schmidt et al.23, the larger scanning range of the IOS for crown models leads to an increase in the number of images, which, in turn, causes image-stitching distortion. Therefore, a higher-crown model corresponds to a larger scanning range and lower accuracy. For inlay specimens, Khaled et al.24 prepared inlays with gingival floor depths of 3 and 4 mm and found that the preparation depth had no significant effect on scanning accuracy. This suggests that within a certain depth range, the scanner’s ability to capture accurate data may not be compromised by depth variations. However, Park et al.25 observed the opposite, noting that precision decreased with increasing inlay depth, with depths tested from 1 to 2 mm. These discrepancies may stem from differences in the inlay depths, shapes, or experimental conditions. In the present study, inlay depth had no significant impact on scanning precision, which remained within clinically acceptable thresholds for all results. This is likely because the depths tested in this study were all within the focal depth range of the intraoral scanner. However, this does not imply that deeper inlays do not pose challenges. In fact, deeper areas showed greater deviations in the 3D fitting analysis, indicating that depth can affect scanning precision. This aligns with the findings of Emam et al.26, who showed that depth variations can influence scanning accuracy, emphasizing the need to consider the impact of depth in dental scanning. Furthermore, inlays with complex geometries can lead to increased deviations in scanning accuracy. This is due to intensified light reflection, attenuation, and scattering, which are particularly problematic in deep cavities and narrow channels. These optical phenomena disrupt light reflection and hinder the scanner’s ability to capture accurate data27.
The accuracy of both indirect digital impression techniques was lower than that of the direct digital impression technique. The results of this study indicated that EOS-S exhibited higher accuracy than EOS-A for both crowns and inlays. German et al.28 demonstrated that the detail reproduction ability of impression materials is closely related to their rheological properties. Specifically, higher fluidity is correlated with improved detail reproduction ability17. Silicone rubber exhibits superior fluidity compared with alginate, which is more susceptible to alterations in impression dimensions owing to factors such as dehydration shrinkage, evaporation, and water absorption29. This may explain the better trueness observed in the EOS-S group. Figure 2 demonstrates that both EOS-A and EOS-S follow a similar trueness trend: for “h”, an increase in diameter at a constant height leads to an increase in the value of trueness. Furthermore, Fig. 4 indicates that crowns of identical height but larger diameters exhibit greater deviation. This phenomenon can be attributed to the increased contact area between the impression material and the model, which results in a stronger hindering effect when the model is removed, leading to a dimensional change in height. In addition, silicone rubber exhibits superior stress tolerance and elastic recovery compared with alginate19,20. This explains why the EOS-S group was less affected by diameter size than the EOS-A group and why the EOS-S group demonstrated fewer deviation for both crowns and inlays, as illustrated in Fig. 4. For “ra” and “rb,” specimens with smaller diameters exhibited worse trueness at the same height. Notably, crown specimens created using the indirect digital impression technique were derived by scanning the recessed areas. The EOS generates a virtual 3D surface by projecting light onto the impression surface, with the reflected light captured by a sensor. However, a reduced diameter can obstruct light projection and reflection because of the narrow lumen, potentially resulting in inaccurate digitization of points30. Furthermore, the properties of the impression material, including fluidity and viscosity, critically influence its accuracy. These properties may limit the material’s ability to flow into narrow areas, thereby impairing its ability to capture fine details, and ultimately compromising impression precision20.
Similarly, for inlays, the data obtained from indirect digital impression techniques were derived by scanning the raised areas of the specimens. As illustrated in Fig. 4, the EOS-A group exhibited a yellow deviation on the sidewalls and a blue deviation on the bottom surface. During the solidification of alginate, pressure applied to the tray may induce stress, which, when released upon model removal, can lead to dimensional changes in the impression18. In this study, the failure rate of inlay impressions was higher than that of crown impressions. Most failures occurred at the fracture point of the raised portion of the negative mold, where it connected to the bottom surface. This can be attributed to the properties of alginate, a viscoelastic material with rubber-like pliability that adheres tightly to the surface of the models, making its removal challenging. Even in the absence of fractures, the raised section of the negative mold tended to shift from its original position owing to suction, resulting in misalignment.
The results of this study indicated that the relevant errors (precision) were within clinical thresholds for all groups, except for the rd index of A1 in the inlay specimens of the EOS-A group, which slightly exceeded the clinical threshold. This finding suggests that variations in the dimensions of crowns and inlays do not significantly affect the digital reproducibility of the impression techniques. However, no significant correlation was observed between precision and trueness. For example, the rb index of C1 in the crown specimens of the EOS-A group exhibited trueness well above the clinical threshold; however, its precision was measured at 0.001792. This indicates that the discrepancies between the specimen and standard model were not owing to human factors, such as operator-related errors in the impression or scanning process.
The alignment process in Geomagic Control X software employs the least-squares method. First, the centroid of the point cloud is calculated based on the geometric features of the model. Subsequently, the macroscopic orientation of the model is determined through translation and rotation in space by aligning the measured model with the reference model. This approach maximizes the overlap of corresponding microscopic point clouds to ensure optimal alignment. However, for regular geometries with symmetric and repetitive characteristics, these alignment methods may fail to accurately identify and align key features, resulting in misfitting between the surfaces. To address this, a 2-mm notch was introduced in the design of the model used in this study. In the initial alignment phase, the software efficiently determines the spatial orientation of the model through rapid identification of notch positions and morphological characteristics. During the subsequent best-fit phase, the three-dimensional coordinate data of the notches significantly enhance the precision of point cloud matching calculations when employing the least squares method. This computational approach ensures rigorous directional alignment and precise positional correspondence between the experimental model and the reference model within the three-dimensional coordinate system. This design contrasts with the regular model used in previous studies31.
Furthermore, when employing the commonly used fitting error ranges of − 0.05 to 0.05 and − 0.1 to 0.132,33,34, most areas appeared green without significant bias. To capture and analyze smaller errors more sensitively, the fitting error range was reduced to −0.01 to 0.01 in this study. Notably, the 3D comparison results reflected the maximum deviation observed across each part of the 10 samples. Although these results provide valuable insights, they do not represent the absolute accuracy of any individual specimen34.
The present study has some limitations. For instance, the in vitro methods used cannot fully replicate clinical factors such as saliva, tongue, and lips, or environmental conditions such as light, temperature, and humidity35,36. Additionally, this study focused on single-tooth preparations, which may not fully represent clinical scanning conditions where neighboring structures could affect accessibility or introduce scanning artifacts. The stainless steel standard model differs significantly from natural dentin and enamel in terms of surface characteristics, including roughness and light reflectivity. Furthermore, this study focused on one IOS, one EOS, and one type each of alginate and silicone. Future research should explore various scanner models and a broader range of impression materials to improve the reliability of these findings.