الفهرس 
SummaryOnly 14 pages are availabe for public view
 |  | from | 116 |  |  |
| | | from | 116 | | |
Abstract
Magnetic resonance imaging (MRI) for brain tumor segmentation is important for diagnosis, estimation of growth rate, measurements of tumor volume and brain tumor treatment preparation. The difficulties of segmenting brain tumors are primarily due to high differences in the size, form, regularity, position and heterogeneous appearance of brain tumors. Automatic brain tumor segmentation is a promising area of research due to recent developments in deep convolutional neural networks (CNN) for semantic image segmentation. Therefore, early-stage identification is very critical in care for increasing the life expectancy of patients. Automatic segmentation of the images would also greatly reduce the burden and also enhance the diagnostic process of the tumors. Several deep convolutional network architectures are suggested that have been effective in segmenting semantic and medical images. Objectives of this work are developing algorithms that segment MRI images into: (ED), (ET), (BG), and (NCR/NET) with an emphasis on increasing accuracy, automatic brain tumor segmentation compared to traditional methods is considered to be less time consuming as it saves doctors’ time and efforts and compete in relation to performance metrics with state-of-the-art methods. well-known architectures, i.e., U-Net, VGG16-Segnet, and DeepLabv3+ models, are applied for the task of brain tumor segmentation. All experiments are applied based on MICCAI’2018 (HGG) subset consisting of 210 brain T1c MRI volumes, each of 155 cross-sections. The proposed method can achieve DSCs of 0. 923, 0. 936 and 0. 961 for the segmentation of the ET, TC, and WT, respectively. Comparison with related works confirm the promise of the proposed methods.