| dc.description.abstract | Magnetic Resonance Imaging (MRI) is widely used in the diagnostic and treatment evaluation of brain tumors. Segmentation is a critical step of the tumor assessment, which usually is a time-consuming task by conventional image analysis methods. In this thesis, I utilized deep learning methods to automate the tumor segmentation and classification tasks. Two models were used, a segmentation model and a classification model. I used U-Net for the segmentation task and a Convolutional Neural Network followed by fully connected layers for the classification task. I evaluated networks on the Mul- timodal Brain Tumor Segmentation Challenge 2020 (BraTS 2020) dataset. Image slices were sampled from the axial axis using three modalities, T1- Contrast-Enhanced, T2-weighted, and Fluid-attenuated inversion recovery. 2-dimensional image slices were used for training in the segmentation task, and annotated images were used for training during the classification task. | |
