Acceptability and deviation of finish line detection and restoration contour design in single-unit crown: Comparative evaluation between 2 AI-based CAD software programs and dental laboratory technicians

Accurate finish line detection and restoration contour design are critical for the clinical success of fixed dental prostheses. While fully automated artificial intelligence (AI)-based computer-aided design (CAD) software programs have demonstrated potential, their virtual design's acceptability and deviation compared with conventional methods remain unclear. The purpose of this in vitro study was to compare the acceptability and deviation of finish line detection and virtual restoration design between 2 fully automated AI-based CAD software programs and dental laboratory technicians. Digital scans of 100 natural abutments prepared for single crowns were replicated 3 times and assigned to dental laboratory technicians (DT), Dentbird (DB), and Automate (AM). Restoration designs were assessed qualitatively by 6 prosthodontists for acceptability and quantitatively using deviation metrics, including root mean square (RMS) error and the Hausdorff distance (HD). Statistical analyses included ANOVA and Student t tests to evaluate intergroup differences (α=.05). Both fully automated systems successfully completed most restorations, with success rates of 97% for DB and 99% for AM. The DT and AM groups demonstrated significantly higher acceptability scores for finish line detection and restoration design than the DB group (P<.001). Quantitative analysis revealed that AM restorations exhibited lower deviations in both RMS values (184 ±36 µm) and HD (132 ±57 µm) than DB, aligning with virtual design acceptability assessments. The Automate program exhibited an acceptability score comparable with that of dental laboratory technicians in finish line detection and restoration design, as well as significantly lower deviation than the Dentbird program.