A 3D Clinical Face Phenotype Space of Genetic Syndromes using a Triplet-Based Singular Geometric Autoencoder

Abstract

AbstractClinical diagnosis of syndromes benefits strongly from objective facial phenotyping. This study investigates facial dysmorphism of genetic syndromes by building and investigating a low-dimensional metric space referred to as the clinical face phenotypic space (CFPS). As a facial matching tool for clinical genetics, such CFPS can enhance clinical diagnosis. It helps to interpret facial dysmorphisms of a subject by placing them within the space of known dysmorphisms. In this paper, a triplet loss-based autoencoder developed by geometric deep learning (GDL) is trained using multi-task learning, which combines supervised and unsupervised learning approaches. Experiments are designed to illustrate the following properties of CFPSs that can aid clinicians in narrowing down their search space: A CFPS can 1) classify and cluster syndromes accurately, 2) generalize to novel syndromes, and 3) preserve the relatedness of genetic diseases, meaning that clusters of phenotypically similar disorders reflect functional relationships between genes. This model is composed of three main components: 1) an encoder based on GDL that optimize distances between individuals in the CFPS therefore adding to the classifier’s power. 2) a decoder that improves both classification and clustering performance by reconstructing a face from an embedding in a CFPS, 3) a singular value decomposition layer to maintain orthogonality and optimal variance distribution across dimensions. This allows for the selection of an optimal number of CFPS dimensions as well as improving the classification, reconstruction, and generalization capabilities of the CFPS.