Identification of Brain Tumor using Deep Learning Neural Networks

Locating the brain tumor is the tough task in field of the medical. In this research work, an Convolutional Neural Network (CNN) is developed based on auto context based on the global and local image features of two dimensional patches with different sizes. Different types of models are considered: 1) a voxel wise technique derived from two dimensional pathways with different direction (coronal, sagittal and axial) and also used three dimensional image without using expensive convolution networks: 2) A convolution network that has been fully automated based on principle of U-net architecture. To locate the boundaries of tumor in brain and other regions posterior probability maps is used via through the networks are used as context information and compared with the original image.