Diagnosis of rare diseases based on facial phenotype: a quantitative assessment using 2D and 3D photography in Stickler syndrome

Abstract Stickler syndrome (SS) is a rare genetic disorder caused by mutations in collagen-encoding genes. Patients with SS are at an increased risk of retinal detachment, with ophthalmologic complications often presenting as the first symptom due to significant diagnostic delays. These delays contribute to severe visual impairment, even in early childhood, making diagnostic uncertainty a major clinical challenge. This study assessed the feasibility of early SS detection using facial features. Supervised machine learning models, including XGBoost and support vector machines (SVM), were applied to 2D and 3D facial photographs to differentiate individuals with SS from healthy controls. The models achieved 92% accuracy in identifying SS patients using full-face 2D and 3D photographs. When focusing specifically on the orbitonasal region in 3D images, classification accuracy increased to 96%. Facial morphology in rare diseases such as Stickler syndrome can serve as a highly accurate, non-invasive screening tool. AI-based analysis has the potential to reduce diagnostic delays and prevent associated complications, improving patient outcomes.