Late Prediction-Fusion Approach for Multimodal Brain Tumor Segmentation

INST: L_042MRI scans play an important role in the examination of brain tumors, including segmentation. Since the tumors are heterogeneous either in shape, size and location, developing an algorithm that automatically segments brain tumors is a very challenging task. Deep learning methods, specifically convolutional networks have achieved outstandingly good results in this area. We achieve Dice coefficients of 0.900, 0.896 and 0.830 for the whole tumor, the tumor core and the enhancing tumor respectively