Brain Region Segmentation using Convolutional Neural Network

Brain region segmentation or skull stripping is an essential step in neuroimaging application such as surgical, surface reconstruction, image registration etc. The accuracy of all existing methods depends on the registration and image geometry. When this fails, the probability of success is very less. In order to avoid this, Convolutional Neural Network (CNN) is used. For brain extraction which is free from geometry and registration. CNN learned the connectedness and shape of the brain. OASIS database is used which is publicly available benchmark dataset. In this method, training phase uses 30 images and 10 images are used for testing phase. The performance of CNN results is closer to the ground truth results given by experts.