Artificial Intelligence-Driven Automated Design of Anterior and Posterior Crowns Under Diverse Occlusal Scenarios

To evaluate the impact of occlusion type and artificial intelligence-based computer-aided design (CAD) software on the geometric accuracy and clinical quality of auto-generated anterior and posterior crown designs.

Five typodont models representing various occlusion types (normal, Class I anterior diastema, Class II division 1, Class II division 2, and Class III anterior crossbite occlusion) underwent crown preparation for the maxillary right central incisor and first molar. Ten sets of intraoral scans were obtained from each prepared model, and crown designs were automatically generated using two software programs: deep learning-based (DL; Dentbird) and conventional automated (CA; Auto Workflow, 3Shape) (n = 10). Surface deviations between the crown designs and preoperative tooth morphology were quantified using root mean square (RMS) values. Clinical crown quality was assessed using World Dental Federation (FDI) criteria. Scheirer-Ray-Hare and Fisher's exact tests were conducted (α = 0.05).

Significant differences in surface deviation and clinical quality were observed between the various occlusion and software types. The DL group demonstrated higher RMS values than the CA group (p < 0.001). However, DL-generated crowns were of significantly better clinical quality (FDI scores) than CA-generated crowns, particularly for posterior teeth, in terms of marginal adaptation, proximal contacts, and anatomical form and contour (p < 0.05). The DL group demonstrated generally favorable outcomes when designing crowns for normal occlusion, but outcomes were less satisfactory when designing anterior crowns with diastemas.

Occlusal scenarios influenced the surface deviation and quality of auto-generated anterior and posterior crown designs. DL software produced higher-quality molar designs than CA software.

Automated crown design outcomes depend on occlusal scenarios and CAD software selection. DL-based CAD software demonstrated superior clinical quality, particularly for posterior crowns, indicating higher clinical suitability. However, further software refinement is needed to consistently produce clinically acceptable crowns under diverse occlusal conditions, such as anterior diastemas.