E1D3 U-Net for Brain Tumor Segmentation: Submission to the RSNA-ASNR-MICCAI BraTS 2021 challenge

Convolutional Neural Networks (CNNs) have demonstrated state-of-the-art performance in medical image segmentation tasks. A common feature most top-performing CNNs is an encoder-decoder architecture inspired by the U-Net. For multi-region brain tumor segmentation, 3D U-Net and its variants provide competitive performances. In this work, we propose interesting extension of standard architecture, specialized for segmentation. The proposed network, called E1D3 U-Net, a one-encoder, three-decoder fully-convolutional neural network where each decoder segments one hierarchical regions interest: whole tumor, core, enhancing core. On BraTS 2018 validation (unseen) dataset, demonstrates single-prediction comparable with networks reasonable computational requirements without ensembling. As submission to RSNA-ASNR-MICCAI 2021 challenge, also evaluate our proposal on dataset. showcases flexibility which exploit task