Instance segmentation plays a vital role in digital pathology image analysis. However, challenges such as low contrast, structural adhesion, and background noise are still common in medical images. These often result in under-segmentation, over-segmentation, and misclassification, which hinder the robustness and accuracy of current methods. To address these issues, we propose SW-BiMambaNet, an efficient and context-enhanced method for nuclei instance segmentation. Specifically, we design a state-space modeling encoder, SW-BiMamba, which combines a shifted window mechanism with a bidirectional state propagation strategy. This design achieves an effective trade-off between computational costs and representational capacity, enabling the joint modeling of global context and local details, which is crucial for accurately distinguishing densely clustered and background-obscured nuclei. Furthermore, we introduce a Hierarchical Cross Attention Fusion (HCAF) module to integrate multi-scale semantic and cross-level structural information. This design further improves the model’s ability to distinguish low-contrast targets and adherent nuclei. Extensive experiments on the MoNuSeg and PanNuke datasets demonstrate that SW-BiMambaNet combines competitive accuracy with significantly lower computational costs. The source code will be publicly available at https://github.com/Oliviachanly/SW-BiMambaNet.git .