AI thesis regarding syndrome vs. control classification of faces of people with Intellectual Disability.
Fien Ockers, March-August 2020
Project for my master AI thesis regarding syndrome vs. control classification of faces of people with Intellectual Disability.
Abstract of paper:
Diagnosing a genetic syndrome can be difficult as extracted genetic data is not always decisive. To reach a diagnosis,
doctors can compare the facial characteristics of a patient with previously diagnosed patients. This process is subjective
and therefore doctors would benefit from an objective model to compare these faces. A Hybrid model has been created at the
Radboudumc that creates face representations. This research aims to find a model that performs better than the Hybrid model
for face classification. Four different models have been used for syndrome vs. control classification, including one ensemble
model. The three other models use either a face representation based on a neural network, a 3D landmark representation or a
morphometric representation based on these 3D landmarks. However, none of them had a better overall performance than the
Hybrid model. Future research can focus on rewriting the Hybrid model, to reduce computation time and make it usable.