Introduction:
Accurate detection of the finish line is critical for the marginal fit and clinical success of dental restorations. Previous studies have shown that AI-driven and deep learning-based systems can reliably detect finish lines1,2, but their performance may be influenced by margin geometry and software choice.
This study aims to evaluate how different finish line designs affect the accuracy of automatic margin detection using various AI-powered dental CAD software platforms
Purpose:
The aim of this study is to evaluate the impact of finish line design on the accuracy of automatic margin detection by AI softwares
Methodology:
•Four tooth preparations with different finish line designs were created digitally:
• Feather edge (0.3 mm),
• Chamfer (0.8 mm),
• Deep chamfer (1.0 mm),
• Shoulder (1.0 mm),
•The models were printed and scans were acquired with TRIOS 4 (3Shape A/S).
•Each preparation was analyzed using two dental CAD software platforms: Atomica AI (Atomica Dental Technology) and 3Shape Automate (3Shape A/S) to automatically detect the finish line margin.
•A reference standard margin for each preparation was established by the researcher during the crown design process in Dental System 2024 (3shape A/S).
•The accuracy of the AI-CAD-detected margins was evaluated by comparing them to the reference margins.
•Geomagic was used to calculate the deviation between the detected and reference finish line margins.
results:
•Two-way ANOVA was used to evaluate main effects of finish line type and software, and interaction effect between finish line type and software.
•RMS values showed no significant differences among the different finish line designs.
•No significant differences were found between the different softwares (p>0.05).
Results:
•Two-way ANOVA was used to evaluate main effects of finish line type and software, and interaction effect between finish line type and software.
•RMS values showed no significant differences among the different finish line designs.
•No significant differences were found between the different softwares (p>0.05).
Conclusions:
It was hypothesized that margin detection would be less accurate for the feather margin design. However, there was no significant difference in margin detection across the different finish line designs. There was also no significant difference between the different softwares and the reference margins. It is believed that due to the crown margins being designed digitally they were very clear and easy to detect. In conclusion, AI margin detection is an adequate method for crown design.
References
1.Mai, Hoa Thi, Jae-Won Han, Yun-Hoa Kim, Myoung-Uk Jin, and Jeong-Yol Lee. "Reliability of Automatic Finish Line Detection for Tooth Preparation in Dental Computer-Aided Software." Journal of Prosthodontic Research, vol. 67, no. 1, 2023, pp. 110–117. https://doi.org/10.2186/jpr.JPR_D_21_00344 Choi, Da-Hye, Seung-Ho Baek, Tae-Ho Kim, and
2.Ji-Man Park. "Deep Learning-Based Automated Detection of the Dental Crown Finish Line: An Accuracy Study." Scientific Reports, vol. 13, no. 1, 2023, article no. 21986. https://doi.org/10.1038/s41598-023-48783-x.
