Context-aware network with enhanced local information for medical image segmentation

• Published in: Pattern analysis and applications : PAA Vol. 28; no. 3
• Main Authors: Liu, Shangwang, Wang, Hongwei, Lin, Yinghai, Jin, Xianglian, Wang, Yusen, Cheng, Yulin
• Format: Journal Article
• Language: English
• Published: London Springer London 01.09.2025; Springer Nature B.V
• Subjects: Artificial neural networks; Attention; Complexity; Computer Science; Datasets; Decoding; Image segmentation; Medical imaging; Modules; Original Article; Pattern Recognition; MCE module; Attention mechanisms; MFP module; Medical image segmentation
• ISSN: 1433-7541; 1433-755X
• DOI: 10.1007/s10044-025-01496-9

In the field of medical image segmentation, architectures such as Unet and APFormer based on CNN (Convolutional Neural Network) and Transformer have made significant progress. However, they still face some challenges when dealing with complex datasets, such as high computational complexity and insufficient integration of contextual information. To address these issues, we propose a hybrid model for medical image segmentation that combines CNN, attention mechanism and R-MLP. Specifically, based on the coding stage of the U-shaped architecture, we introduce the attention mechanism,and propose the MFP module, which aims to compensate for the loss of global information after the encoder, and to reduce the semantic gap between the low-level features and the high-level features during the encoding and decoding process. In particular, we also propose the MCE module, which further extracts local information after capturing global information at the bottleneck layer. We conducted experiments on BUSI, DDTI, and PH2 datasets, and the results show that our model performs well in terms of F1, IoU, HD95, and ASD metrics, reaching 80.13%, 72.13%, 15.8642 mm, and 5.0325 mm on the BUSI dataset, respectively.Thus, compared with the state-of-the-art approaches, the proposed model shows a significant improvement compared to the state-of-the-art methods. The code is available at https://github.com/wang-xiang223/MARM-UNet .