Morphology and fracture behavior of lithium disilicate dental crowns designed by human and knowledge-based AI.

This study aimed to compare the occlusal morphology and fracture behavior of lithium disilicate ceramic dental crowns on 12 human participants' premolar #45 designed by a knowledge-based AI (CEREC, biogeneric individual function, BI) and different human personnel (experienced technician, TD, and trained dental students, AD) using CAD software. Digital datasets of crown design were best-fit aligned with the original teeth to evaluate profile and volume discrepancies of the occlusal morphology, and difference in the functional cuspal angle. Milled and sintered lithium disilicate crowns were resin-luted to 3D-printed dental casts and were subjected to axial load-to-fracture test. The fracture loads and failure modes were recorded and examined. Repeated measures ANOVA with LSD post-hoc test, Kruskal-Wallis test, Pearson's correlation coefficient, paired t-test, and chi-square exact test were used for statistical analyses (α = 0.05). BI-generated crowns showed the highest occlusal profile discrepancy (0.3677 ± 0.0388 mm), whereas human-CAD designed crowns showed higher conformity to the original teeth (0.3254 ± 0.0515 mm for TD, 0.3571 ± 0.0820 for AD; z-difference method; p < 0.001). Cusp angle values were significantly different in all groups except BI and TD (54.76 ± 3.81° for the original teeth, 70.84 ± 4.31° for BI, 67.45 ± 5.30° for TD, and 62.30 ± 7.92° for AD; p < 0.001). Although all three groups of crown designs could achieve clinically acceptable fracture resistance (1556.09 ± 525.68 N for BI, 1486.00 ± 520.08 N for TD, 1425.77 ± 433.34 for AD; p = 0.505) such that no significant difference in fracture strength was found, most crowns presented catastrophic bulk fracture that was not clinically restorable because of the substrate fracture. Group BI had a significantly higher percentage of restorable substrate damage than TD (p = 0.014) and AD (p < 0.001). In conclusion, in designing lithium disilicate dental crown, CAD design with human may be better than knowledge-based AI.