A Rapid Implant Generator Based on Neural Network

Abstract

The geometric design of an implant significantly controls the success rates of implant surgery and influences osseointegration. The development of additive manufacturing (3D printing) also makes dental implants with complicated structures feasible to be manufactured. However, conventional computer-aided design (CAD) requires many engineering techniques, and it is time-consuming to generate a dental implant by its parametric design methodology. To more efficiently and conveniently generate the dental implant model, this study developed an Encoder–Decoder neural network with a multi-scale of images to be an alternative to a parametric implant generator in CAD. The network successfully generated the geometry of a dental implant by giving 15 geometrical parameters around 150 microseconds, which are the physical features to define a dental implant. In addition, users without any technical background in CAD are also able to design a dental implant through this network. Furthermore, we can incorporate this model with biomechanical evaluations of dental implants, and then the optimized and customized geometries of dental implants can be generated.