A Dual Discriminator Adversarial Learning Approach for Dental Occlusal Surface Reconstruction

Abstract

Objective. Restoring the correct masticatory function of partially edentulous patient is a challenging task primarily due to the complex tooth morphology between individuals. Although some deep learning-based approaches have been proposed for dental restorations, most of them do not consider the influence of dental biological characteristics for the occlusal surface reconstruction. Description. In this article, we propose a novel dual discriminator adversarial learning network to address these challenges. In particular, this network architecture integrates two models: a dilated convolutional-based generative model and a dual global-local discriminative model. While the generative model adopts dilated convolution layers to generate a feature representation that preserves clear tissue structure, the dual discriminative model makes use of two discriminators to jointly distinguish whether the input is real or fake. While the global discriminator focuses on the missing teeth and adjacent teeth to assess whether it is coherent as a whole, the local discriminator aims only at the defective teeth to ensure the local consistency of the generated dental crown. Results. Experiments on 1000 real-world patient dental samples demonstrate the effectiveness of our method. For quantitative comparison, the image quality metrics are used to measure the similarity of the generated occlusal surface, and the root mean square between the generated result and the target crown calculated by our method is 0.114 mm. In qualitative analysis, the proposed approach can generate more reasonable dental biological morphology. Conclusion. The results demonstrate that our method significantly outperforms the state-of-the-art methods in occlusal surface reconstruction. Importantly, the designed occlusal surface has enough anatomical morphology of natural teeth and superior clinical application value.

Multi-Granularity Tooth Analysis via Faster Region-Convolutional Neural Networks for Effective Tooth Detection and Classification

“…Deep learning (DL) has emerged as a powerful approach for overcoming the challenges in the dentistry domain, capable of autonomously extracting high-level and discriminative characteristics from a given dataset [12]. Convolutional neural networks (CNNs) have achieved significant appeal among DL approaches due to their well-established multilayer structure.…”

A data-driven approach for the partial reconstruction of individual human molar teeth using generative deep learning

“…Derived from a 3D mesh object as the primary input format, the standard transformation of the raw data typically entails converting the data into one of two formats. The widespread availability and accessibility of 2D generative network architectures tailored for image inference (Isola et al, 2017 ; Karras et al, 2017 , 2020a , b ; Pandey et al, 2020 ; Ding et al, 2021 ; Lei et al, 2022 ; Tian et al, 2022c ) have prompted the adoption of a 2D depth map representation for the data in various studies (Hwang et al, 2018 ; Yuan et al, 2020 ; Tian et al, 2021 , 2022a , b ).…”

“…Tian et al ( 2022a ) used a similar approach for the reconstruction of full occlusal surfaces with a dilated convolutional-based generative model and a dual global-local discriminative model. The proposed generative model utilizes dilated convolution layers to generate a feature representation that preserves the clear tissue structure, while the dual discriminative model employs two discriminators to jointly assess the input.…”

“…The targeted tooth preparations for restoration were accompanied by two adjacent teeth in the same depth map, along with three antagonists and the gap distance between the two jaws in terminal occlusion (Hwang et al, 2018 ; Yuan et al, 2020 ; Tian et al, 2021 ), each saved as a separate image. Moreover, Tian et al ( 2022a , b ) broadened the scope of input data by integrating extracted dental biological morphology, including features such as the occlusal groove and the occlusal fingerprint (Kullmer et al, 2009 ). The ground truth in these studies consisted of single crowns designed by a technician.…”

“…Moreover, unlike the data quantity integrated by the previously discussed methods (Hwang et al, 2018 ; Yuan et al, 2020 ; Tian et al, 2021 , 2022a , b ), our proposed method utilizes a dataset comprising 92 mesh files of single molar crowns. This showcases the proficiency of the 2D methodology in restoring varying extents of the occlusal surface with a minimal amount of data.…”