3D facial shape analysis: Unraveling environmental, genetic, and health connections through AI-driven methodological advancements

The human face, shaped by genetic and environmental factors, reflects health outcomes and developmental problems. To address the limitations of traditional phenotyping approaches, this thesis develops AI-based methods for 3D facial shape analysis, focusing on data-driven and automatic phenotyping which highlights interpretability and confounder control. Leveraging geometric deep learning, the research introduces pipelines for genetic analysis, epidemiological studies, and health risk assessment, while ensuring data privacy through federated learning. The findings demonstrate AI’s potential to enhance clinical applications of facial phenotyping, offering new insights into the relationships between facial morphology, genetics, and health outcomes.