An ontology-based technique for validation of MRI brain segmentation methods

We propose an ontology and rules based approach as innovative instrument to improve and validate brain segmentation in Magnetic Resonance Imaging (MRI), which is a very difficult and time consuming problem. Different techniques are realized to automate segmentation and their development requires a careful evaluation of precision and accuracy. At present segmentation procedures are generally validated by comparison with brain atlas or by use of phantoms. We combine ontology and rules to formalize knowledge about normal and anomalous distribution of brain tissues. Automatic reasoning points out possible "anomalies", imputable to segmentation procedure: in this way the detection and the subsequent solution of bugs become viable.