Temporal lobe epilepsy (TLE) is the most frequent form of drug-resistant focal epilepsy which is mostly characterized by complex seizures, changes in hippocampal and sulci shapes. Numerous automated segmentation techniques have been proposed to be used in the evaluation of TLE. However, the performance of the available methods is still far from manual segmentation and most of these methods are dedicated to adult’s anatomy. The objective of our project is to develop a morphological predictor for early diagnosis of TLE which can identify differences in hippocampal and sulci shapes to normal populations. An atlas specifically for pediatric population will be created which will enable automatically segmenting the hippocampus, insula and sulci curves. Non-linear manifold embedding will also be used for registration to be able to analyze changes and do early detection of TLE.
