Deep semantic segmentation for MRI brain tumor

Abstract

The Magnetic Resonance Images (MRI) the segmentation of brain tumor is quite possibly the most troublesome medical images segmentation and it has many challenges because the tumor has no specific shape or size, not found in a specific place of the brain and contains three sub-areas (Full Tumor FT, Enhanced Tumor ET and Tumor Core TC). The Manual segmentation is extremely difficult and prone to mistakes. In this research the semantic segmentation is used by applying the U-net model, which is a "Fully Convolutional Network (FCN)" algorithm on BraTS 2018 that contains four modalities (Flair, T2, T1,T1c). The results were promised by segmenting the brain tumor within three sub-areas (FT, ET and TC) and the results evaluated using standard brain tumor segmentation metrics. The proposed system achieves "Mean Dice Similarity Coefficient" metric; It is 0.72, 0.86, 0.75 for ET, FT and TC, respectively. Additionally, the "Median Dice Similarity Coefficient" metric is 0.81, 0.91, 0.85 for ET, FT and TC, respectively.