Segmentation of Cerebral Cortex Based on 3D CNN

Peng Liu

Abstract: Research on the surface of the cerebral cortex is crucial to understanding how the brain works. In this study, a three-dimensional full convolution neural network (3D CNN) based on sample equalization mechanism is proposed for MRI cortical region segmentation. We implement a deeper network structure through a small kernel, and obtain multi-scale information by integrating local and global context information to improve the segmentation accuracy. The algorithm is verified by comprehensive experiments on Mindboggle dataset. The results show that compared with other recent methods, this method improves the segmentation accuracy.

Keywords: Deep Learning, Cerebral Cortex Segmentation, 3D CNN