TransBTS: Multimodal Brain Tumor Segmentation Using Transformer

Transformer, which can benefit from global (long-range) information modeling using self-attention mechanisms, has been successful in natural language processing and 2D image classification recently. However, both local features are crucial for dense prediction tasks, especially 3D medical segmentation. In this paper, we the first time exploit Transformer CNN MRI Brain Tumor Segmentation propose a novel network named TransBTS based on encoder-decoder structure. To capture context information, encoder utilizes to extract volumetric spatial feature maps. Meanwhile, maps reformed elaborately tokens that fed into modeling. The decoder leverages embedded by performs progressive upsampling predict detailed segmentation map. Extensive experimental results BraTS 2019 2020 datasets show achieves comparable or higher than previous state-of-the-art methods brain tumor scans. source code is available at https://github.com/Wenxuan-1119/TransBTS.