A System for Brain Image Segmentation and Classification Based on Three-Dimensional Convolutional Neural Network

We consider the problem of fully automatic  brain tumor segmentation in MR images containing  glioblastomas. We propose a three-Dimensional  Convolutional Neural Network (3D-CNN) approach that  achieves high performance while being extremely  efficient, a balance that existing methods have struggled  to achieve. Our 3D-Brain CNN is formed directly on  raw image modalities and thus learn a characteristic  representation directly from the data. We propose a  new cascading architecture with two pathways that each  model normal details in tumors. Fully exploiting the  convolutional nature of our model also allows us to  segment a complete cerebral image in one minute. In experiments on the 2013 and 2015 BRATS challenge dataset; we exhibit that our approach is among the most powerful methods in the literature, while also being very effective.