Effective and Efficient Medical Image Segmentation with Hierarchical Context Interaction
Year: 2025
Type: conference-paper
Abstract: The U-Net architecture has been extensively used in the field of medical image segmentation. Recent advancements have showcased the potential of incorporating attention-based techniques into U-Net structures. However, the integration of attention mechanisms often leads to a substantial increase in both computational demands and the number of parameters, with only a marginal improvement in the performance. This raises questions regarding the parameter efficiency of introducing attention modules. In this paper, we introduce a novel approach termed hierarchical context interaction (HCI), which is a more parameter-efficient and attention-free module that can be seamlessly integrated into U-Net-based architectures. Experimental results demonstrate that the proposed method achieves state-of-the-art performance on both the Medical Segmentation Decathlon Datasets and the Synapse Datasets, while maintaining a competitive computational cost compared to standard U-Net architectures.
Institution: University of Oxford
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4409264134