Detection of Cranial- Facial Malformations: Towards an Automatic Method

This paper presents an original method to detect malformations in cranial facial 3D images. Thisis a hard task because of the complex structure of the cranium. The method is composed byseveral stages containing essentially a segmentation phase to localize the bone structure, a 3Dimage characterization to delineate automatically geometrical characteristics from thesegmented images. These need to possess a mathematical definition in order to be calculatedand an anatomical pertinence to be representative of the structures’ shape. So, thecharacteristics retained are the crest lines. And finally a registration phase to identify thedeformations place. This method is tested on real and synthetic data and has proved hisperformance.