Mask-embedding and feature-fused network for medical image segmentation

Abstract

Medical image segmentation has a vital role in disease diagnosis and treatment. The feature enhancement module and a mask embedding block for medical image segmentation is proposed. This method utilizes an encoder-decoder architecture with attention mechanism and residual connections to adaptively adjust the importance of each layer of features. The proposed network achieves stronger feature transfer and reconstruction, enhancing multi-scale expressive capabilities and context-awareness иy introducing dense skip connections Experimental results on three datasets demonstrate significant improvements in segmentation accuracy and robustness, particularly in handling segmentation details and boundaries.