Automatic brain tumor classification and segmentation by deep neural networks

The target of this thesis is automatic brain tumor segmentation and classification by deep neural networks. Nowadays, brain tumor cases are an important issue. Magnetic Resonance Imaging (MRI) images are used in diagnosis, follow-up and treatment of brain tumor. The evaluation of MRI images by specialists takes time, which leads to the progression of the disease and can even affect the treatment negatively. Otherwise, specialists can misclassify the brain tumor. Therefore, it is important to automatically segment and classify the brain tumor by MRI images. After the diagnosis of the brain tumor, the tumor grows, shrinks or remains in the follow-up phase by following the MRI images. Thus, automatic monitoring at every stage becomes very important. On the other hand, the deep learning methods will be used for automatic segmentation and classification of the brain tumor. Brain tumors are inconsistent in shape, size and magnitude. In this respect, it is difficult to classify them, and it is thought that deep learning methods might give better results than other segmentation techniques.