A fusion‐attention swin transformer for cardiac MRI image segmentation

Abstract For semantic segmentation of cardiac magnetic resonance image (MRI) with low recognition and high background noise, a fusion‐attention Swin Transformer is proposed based on cognitive science deep learning methods. It has U‐shaped symmetric encoding–decoding structure an attention‐based skip connection. The encoder realizes self‐attention for feature representation the decoder up‐samples global features to corresponding input resolution pixel‐level segmentation. By introducing connection between fusion attention, remote interaction information realized, attention local specific channels enhanced. A public ACDC MRI dataset used experiments. left ventricle, right myocardial layer realized. method performs well small sample dataset, example, pixel accuracy obtained by model 93.68%, Dice coefficient 92.28%, HD 11.18. Compared state‐of‐the‐art models, precision been significantly improved, especially heavily occluded targets.