LGI Net: Enhancing local-global information interaction for medical image segmentation

Linjie Liu, Yan Li, Yanlin Wu, Lili Ren, Guanglei Wang

Abstract: Highlights•Innovatively introduced a Local-Global Attention Fusion (LGAF) module that effectively integrates local and global contextual information. This module supplements fine-grained details when modeling long-distance dependencies between features, thereby enhancing the model's robustness.•Proposed a novel approach for information interaction, which takes input sequences from both convolutional and Transformer components and constructs attention feature maps in a cross-guided manner.•Based on the proposed LGAF, a novel decoder is proposed to iteratively reconstruct feature maps from both the encoder and skip connections, accurately generating semantic segmentation maps.