Poster #21: Hongzhuo Chen MA

FaceKit: a Lightweight Toolkit for Interpretable Facial Phenotyping in Rare Diseases

Hongzhuo Chen, Zhanliang Wang, Florent Pollet, Mian Umair Ahsan, Chunhua Weng, Gamze Gürsoy, Kai Wang

Facial phenotyping plays an important role in diagnosis and clinical characterization of rare genetic diseases, many associated with recognizable craniofacial features. However, traditional approaches for describing facial morphology rely on qualitative clinical observation and free-text descriptions, which are subjective, non-standardized, and difficult to reproduce. Although the Human Phenotype Ontology (HPO) provides controlled terms for describing dysmorphic facial features, these terms are categorical rather than quantitative and may vary by examiner experience. Here, we present FaceKit, a computational framework for quantitative facial phenotyping from frontal facial photographs. FaceKit extracts standardized facial landmark measurements and 120 derived morphological features, then reports feature-level z-scores relative to population reference distributions. To support ancestry-aware phenotyping, we constructed reference models using FairFace as a diverse control population. We evaluated FaceKit using multiple rare disease phenotype resources, including GMDB, RDFace, and an internal clinical cohort. FaceKit also includes synthetic image generation for rare disease model development and data augmentation. We performed privacy evaluation to assess whether synthetic images reveal identifiable information from real patient photographs and could therefore compromise patient privacy. Across disease case studies, FaceKit-derived measurements captured known facial features associated with rare genetic disorders and provided objective support for clinical phenotype interpretation. We further demonstrate that FaceKit can assist human reviewers in generating HPO terms and text-based facial descriptions from patient photographs, improving consistency and reproducibility of phenotype annotation. Together, these results establish FaceKit as a useful tool for rare disease research, quantitative dysmorphology, and AI-assisted phenotyping. By converting facial morphology into standardized, quantitative, and interpretable features, FaceKit has the potential to improve rare disease diagnosis, support genotype-phenotype studies, and enable more reproducible clinical characterization across diverse patient populations. FaceKit is available at https://github.com/WGLab/FaceKit.