We proposed a segmentation model combining vision transformers and dense skip connection with attention gate to identify acute stroke lesions from multi-modal MRI scans. This approach achieves state-of-the-art results and outperformed benchmarks including resUnet and seg- former.
