Limitations of RMS error in evaluating intraoral scanner accuracy: Comparison with linear and angular distortions from an enhanced metrology protocol

Several studies have assessed the performance of various intraoral scanners (IOSs) and scan patterns for completely edentulous implant scans. However, cross-referencing the results is challenging because of discrepancies in the superimposition methods and outcome measures. The frequent reporting of a single root mean square (RMS) error value further limits the assessment of complex 3-dimensional (3D) distortions. Consequently, the accuracy of IOSs for completely edentulous implant scans remains unknown.

The purpose of this study was to compare the accuracy of 3 IOSs and their recommended scan patterns on a 5-implant edentulous mandible model. An enhanced metrology software protocol was used and compared with the predominantly applied protocol of best fit alignment (BFA) and RMS error.

A 5-implant edentulous mandible model was fabricated with varying implant angulations, depths, and interimplant distances to simulate a clinical situation. The model was digitized with a laboratory scanner and used as the reference scan. Using 3 IOSs and their manufacturer's recommended scan patterns (Primescan [Pr, A], TRIOS 5 [Tr, B], and iTero Element 5D Plus [iT, C]), 10 test scans (n=10) were generated for each test group. An enhanced metrology protocol using geometric dimensioning and tolerancing (GD&T) features and reference point system (RPS) alignment was used to quantify linear, angular, and distance distortions in a 3D inspection software program (Geomagic Control X; 3D Systems). RMS values were also generated using BFA and compared. Two-way ANOVA and Tukey HSD were used to compare IOSs, scan patterns, and their interactions, while the Wilcoxon signed-rank test was used to interpret the direction of linear distortion (α=.05).

Pr obtained the lowest deviations whenever statistical significance was reached for global linear distortion (d_R), absolute angular distortion (Absdθ), and 3D distance distortion (ΔR), while Tr and iT could not be consistently ranked (P<.05). The interaction effect was not significant for all primary outcomes (P>.05). Scan pattern C showed significantly higher Absdθ than A at implant P (P=.020) and B at implants P (P=.009) and S (P=.001). There was an increase in arch width for Pr and Tr scans and a decrease in arch length on the left for Pr and iT scans. Tr scans showed a wave-like distortion pattern along the z-axis across the arch. RMS ranged from 22.4 ±3.1 μm (Pr_C) to 53.6 ±7.9 μm (Tr_A), with Pr having the lowest RMS, followed by iT and Tr (all P<.001).

The enhanced metrology protocol generated linear, angular, and distance distortion values which allowed for detailed 3D distortion analysis. RMS values were not fully representative of linear and angular distortions. While Pr exhibited the lowest linear, angular, and distance distortions, Tr and iT could not be consistently ranked. BFA-derived RMS values may oversimplify 3D distortions and lead to erroneous conclusions regarding the accuracies of IOSs.