The Deep Convolutional Network Segmentation And Visualization Of MRI Brain Image

The medical imaging has played an important role in medical diagnosis with the development of various medical imaging technologies. It provides scientific proof for doctors to make the decisions. The development of imaging technology produces higher resolution and multi-function images, however exploring valuable data hidden in imaging which makes diagnosis more convenient has been attracted a variety of attentions. The auto-segmentation and three-dimensional visualization of imaging are part of these efforts.Nowdays, lots of image segmentation methods which include many classic ones have been proposed and perform well. The three-dimensional visualization can transform two-dimensional imaging to three-dimensional imaging, so it is widely used in medical domain.The main work in this paper is to build a deep convolutional network, segment the MRI brain imaging by it and then show the stereo three-dimensional reconstructed imaging.First, the paper describes the development of MRI and deep learning, and shows the importance of MRI segmentation. Furthermore, a deep convolutional network architecture used to segment MRI brain imaging is proposed based on the available ones and built by Caffe. Its performance is assessed. To show the stereo results, the methods of stereo display, mainly about color separation display is summarized. Finally, The open-source ImageJ and Volume Viewer are reprogrammed to realize the red-cyan stereo display. The stereo reconstructed imaging is showed by them.