Partial Fixed Dental Prosthesis for Additive Manufacturing: Automated Build Orientation Selection and Optimization Based on Deep Learning

Additive Manufacturing (AM) plays a critical role in producing complex shapes. However, in the current stage, technicians often use their expertise to determine optimal printing orientations. Consequently, the pre-processing for finding the best printing orientations for each model becomes time-consuming. Thus, we proposed a three-stage deep learning framework to optimize the printing orientations of partially fixed dental prostheses. The deep hierarchical feature learning was used to enhance training capabilities, aiding in partial fixed dental prosthesis with single or multiple teeth recognition. Additionally, multi-scale local contextual features on partial fixed dental prostheses were extracted hierarchically. Furthermore, we employed a genetic algorithm to consider two objective functions: surface area and printing time. The results showed candidate orientations. Users can select the most favorable orientation based on the importance of these objective functions. This approach enhanced partial fixed dental prosthesis precision, reduced post-processing time, and accelerated the manufacturing of partial fixed dental prostheses.