3D Semantic Segmentation of Brain Tumor for Overall Survival Prediction

Glioma, a malignant brain tumor, requires immediate treatment to improve the survival of patients. The heterogeneous nature Glioma makes segmentation difficult, especially for sub-regions like necrosis, enhancing non-enhancing and edema. Deep neural networks full convolution an ensemble fully are successful segmentation. paper demonstrates use 3D network with three-layer encoder-decoder approach. dense connections within layer help in diversified feature learning. takes patches from $$T_{1}$$ , $$T_{2}$$ $$T_{1}c$$ FLAIR modalities as input. loss function combines dice focal functions. Dice similarity coefficient training validation set is 0.88, 0.83, 0.78 0.87, 0.75, 0.76 whole tumor core respectively. achieves comparable performance other state-of-the-art approaches. random forest regressor trains on shape, volumetric, age features extracted ground truth overall prediction. accuracy 56.8% 51.7% sets.