Related Experiment Video
Updated: Jan 9, 2026
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
Acceptability, deviation, and efficiency of 2 fully automated CAD software programs in designing 3-unit fixed dental
Kedith Sawangsri1, Mariam Bekkali2, Yuan-Lynn Hsieh1
1Clinical Assistant Professor, Division of Restorative and Prosthetic Dentistry, The Ohio State University College of Dentistry, Columbus, Ohio.
Statement Of Problem:
Fully automated artificial intelligence (AI)-based computer-aided design (CAD) software programs have shown promise for single-unit crown design, but their accuracy, acceptability, and efficiency in designing multi-unit fixed dental prostheses (FDPs) remain unclear.
Purpose:
This in vitro study evaluated the design acceptability, trueness, and efficiency of 2 commercially available fully automated AI -based CAD software programs in designing anterior and posterior 3-unit FDPs compared with those designed by dental laboratory technicians.
Material And Methods:
Digital scans of natural abutments prepared for 3-unit fixed dental prostheses (12 anterior and 18 posterior FDPs) were replicated into 3 sets and allocated for restoration design by Certified Dental Technician (CDT), dentbird (DB), and Atomica AI (AA). Restoration designs were evaluated qualitatively using an 11-criteria acceptability score by 3 calibrated prosthodontists and quantitatively using root mean square (RMS) deviation analysis. CAD times were recorded for each group. Statistical analyses included the Kruskal-Wallis test followed by the Dunn post hoc test, paired t test, and repeated-measure ANOVA followed by the paired t test to evaluate intergroup differences (α=.05).
Results:
DB successfully generated all designs (100%), while AA achieved a 93% overall success rate (83% for anterior and 100% posterior FDPs). AA obtained acceptability scores comparable with those of CDT, particularly in posterior FDPs, while DB exhibited significantly lower scores, especially in anterior FPDs. Both AI programs scored lower than CDT in occlusal and proximal contact areas and connector size (P<.001). AA demonstrated lower RMS deviation than DB in anterior FDPs (P<.05). Both AI systems significantly reduced design time, with AA completing design faster than DB in anterior FDPs (P<.05).
Conclusions:
Fully automated AI-based CAD systems demonstrated variable performance in 3-unit FDP design. AA achieved acceptable trueness and morphology, particularly in posterior regions, while DB underperformed in esthetic and functional areas. Although both systems improved design efficiency, human oversight remains essential for occlusal and proximal contact, connector design, and anterior FDPs.
