Automated micro-CT quantification of clear aligner fit: a pilot comparison of manufacturing processes

Clear aligner fit accuracy significantly influences treatment efficacy. Micro-computed tomography (micro-CT) enables precise three-dimensional quantification of aligner-tooth adaptation, but manual measurement is time-intensive and operator-dependent. To develop and validate an automated flood-fill algorithm for micro-CT gap quantification and compare fit accuracy across three aligner manufacturing processes: in-house 3D-printing, Invisalign (thermoformed), and Angelalign (thermoformed). This in vitro pilot study used a standardized maxillary dental model to fabricate nine aligners across three manufacturing groups (three aligners per group): direct DLP 3D-printing (LuxCreo Inc., Chicago, US), Invisalign (SmartTrack™), and Angelalign (Angel Pro™). Each aligner was scanned using a Xradia 610 Versa micro-CT (ZEISS, Oberkochen, Germany) at 58.82 μm voxel resolution. An automated flood-fill algorithm was developed in Python to segment and quantify aligner-tooth gap widths from reconstructed volumes, and validated against manual segmentation using Dice similarity coefficient, sensitivity, and specificity. Group differences in gap width were assessed using the Kruskal-Wallis test with Mann-Whitney post-hoc comparisons and Bonferroni correction. The automated algorithm achieved a mean Dice similarity coefficient of 0.87 ± 0.04 (range: 0.81–0.94), sensitivity of 0.89 ± 0.05, and specificity of 0.96 ± 0.03. A total of 120 gap width measurements were obtained (3D-printing: n = 42, Invisalign: n = 42, Angelalign: n = 36). Mean gap widths were 193.27 ± 93.03 μm, 186.96 ± 81.86 μm, and 254.89 ± 133.21 μm, respectively. The Kruskal-Wallis test showed no significant overall difference among groups (H = 4.396, P = 0.111). Post-hoc analysis revealed no significant difference between 3D-printing and Invisalign (U = 886.0, P = 0.974, Cliff’s Delta = 0.005, negligible effect). Neither comparison with Angelalign reached statistical significance after Bonferroni correction ( P = 0.070 and P = 0.065), though both showed small effect sizes (Cliff’s Delta ≈ − 0.23). Linear mixed-effects models confirmed these findings (random-effects ICC = 0.179 for aligner-level clustering). The automated flood-fill algorithm may enable efficient and reproducible micro-CT gap quantification, reducing processing time from ~ 45–60 min to approximately 3.2 min per scan. No statistically significant difference in fit accuracy was observed between in-house 3D-printed and Invisalign thermoformed aligners (mean difference: 6.31 μm), providing preliminary evidence for their potential as a cost-effective manufacturing alternative; however, this should not be interpreted as evidence of equivalence given the small aligner sample size. The Angelalign group showed a trend toward larger gaps with small effect sizes, warranting further investigation. Substantial within-group variability across all manufacturing methods underscores the need for systematic quality control in aligner production.