IAMU-Net: Integrated Attention Mechanisms with Coarse-to-Fine Strategy Based on nnU-Net for Multi-Organ Image Segmentation

Automatic multi-organ segmentation plays a critical role in medical image analysis, particularly in preoperative planning and disease assessment. However, conventional segmentation methods often struggle with challenges in handling small organs, ambiguous boundaries, and complex backgrounds, as well as in their generalization ability across different datasets. To address these limitations, we propose a novel model named IAMU-Net for multi-organ segmentation of abdominal computed tomography (CT) scans using deep learning. Our method is built upon the nnU-Net architecture and integrates multiple attention mechanisms to enhance feature representation, spatial information modeling, and multi-scale feature fusion abilities. Furthermore, we use a coarse-to-fine two-stage segmentation strategy to improve the segmentation accuracy of small organs and anatomically complex boundaries. Experimental results on the AbdomenAtlas 1.0 Mini and TotalSegmentator V2 datasets illustrate that our method achieves superior segmentation performance across multiple organs, confirming its effectiveness and robustness.