Medical image segmentation improved by shape aware channel separated network
Morphology-Aware Channel-Decoupled Mamba Network for Medical Image Segmentation
Computer Vision and Pattern Recognition
Summary
Segmenting different organs in CT scans is hard because organs vary in shape and look similar to nearby tissues. The authors propose a new type of neural network that separately processes different parts of data channels and adapts to the local shapes and structures in the images. This approach helps the network better recognize the different organs despite their varying appearances. When tested on a standard dataset, their method showed improved accuracy in identifying multiple organs.
What this means in practice
- •For medical imaging engineers: Create more accurate multi-organ segmentation tools that handle variations in organ shape within CT scans.
- •For radiology software developers: Incorporate advanced local shape adaptation and channel-separated modeling to improve segmentation in diagnostic tools.
Authors
Chenchen Wang, You Zhou
Abstract
Accurate multi-organ segmentation in computed tomography (CT) is essential for computer-aided diagnosis, treatment planning, and disease monitoring. However, substantial variations in organ shape and size, together with low tissue contrast and ambiguous boundaries, complicate the modeling of heterogeneous anatomical structures. To address these challenges, we propose a Morphology-Aware Channel-Decoupled Mamba Network (MCDNet), which integrates adaptive local structural enhancement with channel-decoupled long-range modeling. Specifically, we introduce a Morphology-Aware Channel-Decoupled State-Space (MCSS) block that divides feature channels into separate processing groups, allowing differentiated structural representations to be learned within an efficient state-space framework. Within each group, a Morphology-Aware Enhancement Module (MAEM) combines deformable convolutions with channel attention to adapt local feature extraction to anatomical shape variations. Group-wise state-space modeling then captures long-range spatial dependencies, and channel gating adaptively integrates the resulting features. Experiments on the Synapse abdominal CT dataset demonstrate the effectiveness of MCDNet, achieving an average Dice Similarity Coefficient of 81.61% for multi-organ segmentation.