An Ai-Designed Auxiliary Device For Complete-Arch Intraoral Scanning

To develop auxiliary devices that enhance the accuracy of intraoral scanning for complete-arch implants using a deep-learning AI model. 343 sets of 3D imaging data were collected from a dental laboratory, with 300 sets of complete dental arches used for training and 43 sets of edentulous arches with 4-6 dental implants (28 for validation and 15 for testing). Auxiliary devices, with tooth landmarks placed between implants, were manually designed. An artificial neural network, Multi-Layer Perceptron, was used to predict tooth landmark positions, using normalized implant coordinates as input and landmark coordinates as output. The model was then validated and evaluated using the test group. Test models were digitized using an intraoral scanner under three conditions: 1) without an auxiliary device, 2) with the manual device, and 3) with the AI-generated device. Inter-implant distances and angles were measured and compared. The AI successfully reconstructed auxiliary devices within 10 seconds each. The mean linear accuracy of the AI group is 57.6±43.4 µm, significantly better than the control (77.8±47.5 µm) but not as good as the manual group (38.8±17.2 µm). The mean angular accuracy was 0.145±0.088º, significantly better than the control (0.290±0.102º) and manual group (0.292±0.136º). The AI-generated auxiliary device improves the accuracy of complete-arch intraoral scanning, particularly in reducing angular errors of dental implants. Further research is needed to enhance linear accuracy.