Classification and characterization of facial asymmetry in adult patients with skeletal Class III malocclusion using cluster analysis.

INTRODUCTION: This study aimed to analyze the comprehensive maxillofacial features of patients with skeletal Class III malocclusion and facial asymmetry to develop a classification system for diagnosis and surgical planning. METHODS: A total of 161 adult patients were included, with 121 patients in the asymmetry group (menton deviation >2 mm) and 40 patients in the symmetry group (menton deviation ≤2 mm). Twenty-eight variables were determined, including transverse translation, roll and yaw of each facial unit, transverse width, mandibular morphology, and transverse dental compensation. Principal component (PC) analysis was conducted to extract PCs, and cluster analysis was performed using these components to classify the asymmetry group. A decision tree was constructed on the basis of the clustering results. RESULTS: Six PCs were extracted, explaining 80.622% of the data variability. The asymmetry group was classified into 4 subgroups: (1) atypical type (15.7%) showed an opposite roll direction of maxillary dentition than of menton deviation; (2) compound type (34.71%) demonstrated significant ramus height differences, maxillary roll, and mandibular roll and yaw; (3) mandibular yaw type (44.63%) showed slight mandibular yaw without mandibular morphology asymmetry; and (4) maxillary-shift type (4.96%) shared similarities with the compound type but showed significant maxillary translation. The classification and regression tree model achieved a prediction accuracy of up to 85.11%. CONCLUSIONS: This study identified 4 distinct phenotypes using cluster analysis and proposed tailored treatment recommendations on the basis of their specific characteristics. The classification results emphasized the importance of spatial displacement features, especially mandibular yaw, in diagnosing facial asymmetry. The established classification and regression tree model enables clinicians to identify patients conveniently.

Learning a Prototype Discriminator With RBF for Multimodal Image Synthesis.

Multimodal image synthesis has emerged as a viable solution to the modality missing challenge. Most existing approaches employ softmax-based classifiers to provide modal constraints for the generated models. These methods, however, focus on learning to distinguish inter-domain differences while failing to build intra-domain compactness, resulting in inferior synthetic results. To provide sufficient domain-specific constraint, we hereby introduce a novel prototype discriminator for generative adversarial network (PT-GAN) to effectively estimate the missing or noisy modalities. Different from most previous works, we introduce the Radial Basis Function (RBF) network, endowing the discriminator with domain-specific prototypes, to improve the optimization of generative model. Since the prototype learning extracts more discriminative representation of each domain, and emphasizes intra-domain compactness, it reduces the sensitivity of discriminator to pixel changes in generated images. To address this dilemma, we further propose a reconstructive regularization term which connects the discriminator with the generator, thus enhancing its pixel detectability. To this end, the proposed PT-GAN provides not only consistent domain-specific constraints, but also reasonable uncertainty estimation of generated images with the RBF distance. Experimental results show that our method outperforms the state-of-the-art techniques. The source code will be available at: https://github.com/zhiweibi/PT-GAN.