The paper introduces LKM-UNet, a novel architecture for medical image segmentation that leverages large kernel vision Mamba UNet. The key contributions include: 1. **LKM-UNet Architecture**: A U-shape network that integrates large kernel Mamba modules to enhance both local and global spatial modeling capabilities. 2. **Large Kernel Mamba Blocks**: Novel hierarchical and bidirectional Mamba blocks designed to improve the representation modeling of SSMs, enhancing both pixel-level and patch-level interactions. 3. **Efficiency and Scalability**: Utilizes linear complexity of Mamba to achieve large receptive fields, improving efficiency in local modeling while maintaining superior global modeling capabilities. 4. **Comprehensive Experiments**: Demonstrates the effectiveness of LKM-UNet on 2D and 3D medical image segmentation datasets, showing improved performance over state-of-the-art methods. The paper also includes detailed experimental results, ablation studies, and qualitative segmentation visualizations to validate the effectiveness of the proposed method. The results highlight the importance of large receptive fields and the benefits of bidirectional and hierarchical Mamba designs in medical image segmentation tasks.The paper introduces LKM-UNet, a novel architecture for medical image segmentation that leverages large kernel vision Mamba UNet. The key contributions include: 1. **LKM-UNet Architecture**: A U-shape network that integrates large kernel Mamba modules to enhance both local and global spatial modeling capabilities. 2. **Large Kernel Mamba Blocks**: Novel hierarchical and bidirectional Mamba blocks designed to improve the representation modeling of SSMs, enhancing both pixel-level and patch-level interactions. 3. **Efficiency and Scalability**: Utilizes linear complexity of Mamba to achieve large receptive fields, improving efficiency in local modeling while maintaining superior global modeling capabilities. 4. **Comprehensive Experiments**: Demonstrates the effectiveness of LKM-UNet on 2D and 3D medical image segmentation datasets, showing improved performance over state-of-the-art methods. The paper also includes detailed experimental results, ablation studies, and qualitative segmentation visualizations to validate the effectiveness of the proposed method. The results highlight the importance of large receptive fields and the benefits of bidirectional and hierarchical Mamba designs in medical image segmentation tasks.