Convolutional Neural Network Approach for MR Human Brain Segmentation with U-Net Architecture

Brain segmentation is a potential and essential task for early prediction of diseases. This will further help in treatment planning. The conventional hand segmented brain images are validated unsatisfactory and they are time consuming as well. A substitute methodology is convolutional neural network (CNN) approach with U-Net architecture. 38 volumes of MRI slices were used and validated the performance of this methodology with Dice, Precision, recall and Fl-score. The results show that the model CNNbased U-Net has performed well to predict the data.