Comparison of Artificial Intelligence-Generated and Technician- Designed Maxillary Central Incisor Crowns: An In...

To compare digital designs of maxillary central incisor crowns generated by four artificial intelligence (AI)-based design systems with those created by an experienced dental technician using conventional computer-aided design (CAD) software, with respect to internal surface difference, marginal deviation, incisal and proximal contact characteristics, and design time.

Thirty-two anonymized digital datasets of prepared maxillary central incisors were used. For each prepared tooth, five crowns were generated: one by a master dental technician (MDT) and four using AI-based systems (3shape Automate [AU], Atomica [AT], Dentbird [DT], and Exocad AI Design [EX]). Design time and mesh properties (file size, mesh numbers, and crown volume) were evaluated. Internal surface difference and marginal deviation were assessed using three-dimensional root-mean-square (RMS) deviation analysis. Incisal and proximal contact occurrence and contact characteristics were digitally evaluated. Data were analyzed using repeated-measures ANOVA or the Friedman test, followed by appropriate multiple-comparison procedures (α = 0.05).

Significant differences were identified among groups for design time, file size, mesh numbers, crown volume, internal surface difference, marginal deviation, and proximal contact occurrence (P <0.01). AI-based systems required significantly shorter design times than the technician-designed workflow, with AT showing the lowest values and MDT the highest. The DT and MDT groups exhibited the largest mesh numbers, file sizes and crown volumes. The DT group also demonstrated the smallest internal surface difference (71 ±5 μm), whereas the AT group exhibited the greatest marginal deviation (225.64 ±163.1 μm). Incisal contact occurrence varied among the AI systems and was higher in several groups than in the technician-designed crowns (up to 35.71% vs 7.14%). Significant differences were observed in proximal contact occurrence for both mesial and distal surfaces (P <0.01), and the AT group showed the lowest proximal contact occurrence among all groups (3.57% on both mesial and distal). Despite variability among platforms, all systems produced internal surface difference values within clinically accepted thresholds.

AI-based crown design systems demonstrated variable performance across evaluated parameters for anterior crowns. While certain systems showed favorable internal surface difference values, others demonstrated more balanced incisal and proximal contact behavior. Hybrid workflows combining AI assistance with conventional CAD controls demonstrated comparatively consistent outcomes across parameters. Nevertheless, AI-generated crown designs still require professional evaluation and potential refinement before clinical use.