Deep learning for hippocampus segmentation

Abstract

Hippocampus segmentation plays a key role in diagnosing various brain disorders such as Alzheimer's disease, epilepsy, multiple sclerosis, cancer, depression and others. Nowadays, segmentation is still mainly performed manually by experts. Segmentation done by experts is considered to be a gold-standard when evaluating automated methods, buts it is a time consuming and hard task, requiring specialized personnel.Recently effort has been done to achieve reliable automated segmentation. Currently, the best performing methods are multi atlas based with around 90\% DICE coefficient and very time consuming. Deep learning, a learning technique that is being used in many computer vision applications with great success, is still not very well studied in hippocampus segmentation, with some works recently starting to employ it on this particular segmentation problem.This M.Sc. research project intends to investigate the use of Deep Learning techniques applied to hippocampus segmentation in MR imaging. Our method aims to use the Simultaneous Truth andPerformance Level Estimation (STAPLE) algorithm to generate labeled data as a silver-standard and use it as input for a Convolutional Neural Network (CNN). We will compare our methodology with the state-of-the-art techniques for hippocampus segmentation.