Predicting congenital conditions from facial features

Michelle Terblanche , Riaal Domingues

Partner: fmrg

Year: 2020

Abstract: Manual delineation to determine x-y coordinates of facial features is a resource intensive process. This study evaluated three potential models for the automated detection of facial features to predict two congenital conditions: Mowat-Wilson and Noonan syndrome. Due to very little research being done in the African population, a further goal was to determine the usefulness of the models on African images These automated methods proved viable replacements for manual interventions. However, face detection on African images were less successful and needs further investigation. Results are preliminary and model refinements are required to improve overall stability and predictability.

Presentation Video