Research For Designing Inlay Technology Based On Generative Adversarial Networks

The tooth defect is a frequently-occurring disease and important research direction in the field of dental clinic.The occlusal surface is unnatural and the fit of the prosthesis edge is poor through existing repair methods.Aiming to address the limitations,this paper proposes a network model to restore the tooth crown with personalized anatomical features and fulfilling the full functional requirements.We establish a set of restoration solutions for the tooth with partial defects and large area defects.The main tasks are as following.(1)For the partial defect tooth,the area to be reconstructed is small,and the remaining morphological features is richer.We propose a method for automatic generation of local defect area based on CGAN.Firstly,a crown depth map calculation method based on adaptive visual distance is used to build a training and testing dataset for the network model;Secondly,the network model is trained to balance with the constraints of the opposing tooth and the perceptual adversarial loss to reconstruct the tooth crown with personalized anatomical features.(2)For the tooth with large part defects on occlusal morphology and occlusal relationship,we propose a reconstruction method based on WGAN combined with double discriminators.Firstly,the global and local discriminators are used to calculate the confidence values of the defect tooth and dentition(with adjacent teeth)morphological features,respectively;Secondly,we use the histogram loss to calculate the occlusal spacing distribution,which directly constrains the generator to reconstruct the correct occlusal contact relationship;Finally,the pre-trained occlusal feature extraction network is used to reduce the difference of tooth morphology.(3)Combining the depth map information of the occlusal surface generated by the networks,the reconstruction methods of the inlay and the outside surface,the constraint deformation of the edge shape and the design of the restoration are studied.We restore the point cloud data of the tooth according to the pixel-distance mapping mode and reconstruct the mesh model;Secondly,the inner surface and occlusal surface are segmented,respectively,then the boundary area is adjusted by the constrained deformation algorithm and boundary denoising algorithm;Finally,the boundary between the inner surface and the occlusal surface is connected to design the inlay restoration.