Brain image classification and Segmentation using deep Convolutional neural network

Abstract

The abnormalities in brain cells are the main causes for forming newlinelesions in brain. These abnormal lesions in brain lead to the formation of tumors newlinein brain. Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) newlineare the two different brain image scanning methods. In this research work, MR newlineimages are used to scan the brain internal regions. Benign and Malignant are the newlinetype of abnormal lesions in brain in which, benign can be treated by radiation newlinemethods; whereas malignant lesions are treated through proper surgery by expert newlineradiologist. newlineTumor is defined as an uncontrolled growth of cancerous cells in any newlinepart of the body. Tumors are of different types and possess diverse newlinecharacteristics and require different treatments. At present, brain tumors are newlineclassified into primary brain tumors and metastatic or malignant brain tumors. newlineThe primary tumors begin in the brain and are inclined to stay in the brain; the newlinemetastatic or malignant tumors begin as a cancer elsewhere in the body and then newlinestart to spread into the brain region. Due to the large amount of brain tumor newlineimages that are currently being generated in the clinics, it is not possible for newlinephysicians to manually annotate and segment these images in a practical time. newlineHence, the automatic tumor detection and segmentation technique has become newlineinevitable. In conventional methods, brain tumors are detected and diagnosed newlinemanually by expert radiologist. It is time consuming and error probe process. newlineHence, it is not suitable for high population developing countries. Therefore, a newlinecomputer aided automatic brain tumor detection and diagnosis methods are newlinepreferred. newlineIn this thesis three brain tumor segmentation algorithms are proposed newlineto classify and identify the tumor portions effectively newline