Primary brain tumors Image segmentation based on 3D-UNET with deep supervision and 3D brain modeling

Accurate 3D geometric modeling is essential for surgeons to diagnose and to operate in time. Clear and accurate image segmentation methods are the premise issue of accurate geometric modeling. Due to the complex characters of the brain tissue and the strong function of each part, in this paper, we proposed an image segmentation method based on the 3DU-net neural network with the deep supervision, which rapidly divided primary brain tumors from normal brain tissues. Three-dimensional geometric brain modeling are built so that surgeons may quickly make an accurate diagnose. A simulation platform was built based on Python software, visual studio software, Visualization Toolkit and 3D Slicer software. Experiments results show the accuracy and rapidity of the proposed algorithm.