MRI Brain Tumor Segmentation Using Deep Learning. (Dept. E)

This work presents a method for classification and segmentation of brain tumors based on deep learning analysis contrast T1 (T1c) MR images. To achieve this goal, three different networks are investigated i.e., U-Net, VGG16-Segnet, DeepLabv3+ models. In addition, the integration 3D narrow-band information MRI volumes is imported to input Convolutional Neural Network (CNN) describe more accurately tumor anatomy. Experimentations performed MICCAI’2018 High Grade Glioma (HGG) subset Brain Tumor Segmentation (BraTS) Challenge, composed 210 T1c volumes, each 155 cross-sections. Among CNNs, network achieves highest Dice Similarity Coefficients (DSC) 91.2%, 92.5%, 94.6% Enhancing (ET), Core (TC), Whole (WT), respectively. Comparison with related confirms advantages proposed system.