MOSformer: Momentum encoder-based inter-slice fusion transformer for medical image segmentation

The introduction of MOSformer marks a significant advancement in medical image segmentation, addressing limitations of existing 2.5D-based models that struggle to fuse inter-slice information effectively. By employing dual encoders, MOSformer enhances feature distinguishability across slices, while the inter-slice fusion transformer (IF-Trans) module integrates multi-scale features for improved accuracy. Evaluated on benchmark datasets such as Synapse, ACDC, and AMOS, MOSformer achieved impressive DSC scores of 85.63%, 92.19%, and 85.43%, setting a new standard in the field. This development is crucial for clinical applications that depend on precise medical imaging, potentially leading to better patient outcomes and more efficient diagnostic processes.