LKM-UNet is a novel medical image segmentation model that leverages the Mamba sequence modeling approach to achieve large receptive fields. The model is designed to overcome the limitations of traditional CNNs and Transformers in medical image segmentation, particularly in terms of long-range dependency modeling and computational efficiency. LKM-UNet introduces a Large Kernel Mamba U-shape Network that utilizes large Mamba kernels for efficient spatial modeling, enabling superior performance in both local and global feature modeling. The model incorporates a hierarchical and bidirectional Mamba block to enhance spatial modeling capabilities, allowing for more accurate segmentation of medical images. The key contributions of LKM-UNet include the introduction of a Large Kernel Mamba UNet for 2D and 3D medical image segmentation, the use of large receptive fields in SSM layers to enable large spatial modeling, the design of a bidirectional Mamba for location-aware sequence modeling, and the proposal of a novel hierarchical Mamba module composed of pixel-level and patch-level SSMs. These components work together to improve the model's ability to capture both local and global features in medical images. Experiments on two publicly available datasets, Abdomen CT and Abdomen MR, demonstrate the effectiveness of LKM-UNet in medical image segmentation. The results show that LKM-UNet outperforms existing methods in terms of segmentation accuracy, particularly in capturing detailed local features and global structures. The model's performance is further validated through ablation studies, which confirm the importance of each component in the model's design. The results also highlight the potential of Mamba in modeling both local and global features with larger receptive fields, making it a promising approach for medical image segmentation.LKM-UNet is a novel medical image segmentation model that leverages the Mamba sequence modeling approach to achieve large receptive fields. The model is designed to overcome the limitations of traditional CNNs and Transformers in medical image segmentation, particularly in terms of long-range dependency modeling and computational efficiency. LKM-UNet introduces a Large Kernel Mamba U-shape Network that utilizes large Mamba kernels for efficient spatial modeling, enabling superior performance in both local and global feature modeling. The model incorporates a hierarchical and bidirectional Mamba block to enhance spatial modeling capabilities, allowing for more accurate segmentation of medical images. The key contributions of LKM-UNet include the introduction of a Large Kernel Mamba UNet for 2D and 3D medical image segmentation, the use of large receptive fields in SSM layers to enable large spatial modeling, the design of a bidirectional Mamba for location-aware sequence modeling, and the proposal of a novel hierarchical Mamba module composed of pixel-level and patch-level SSMs. These components work together to improve the model's ability to capture both local and global features in medical images. Experiments on two publicly available datasets, Abdomen CT and Abdomen MR, demonstrate the effectiveness of LKM-UNet in medical image segmentation. The results show that LKM-UNet outperforms existing methods in terms of segmentation accuracy, particularly in capturing detailed local features and global structures. The model's performance is further validated through ablation studies, which confirm the importance of each component in the model's design. The results also highlight the potential of Mamba in modeling both local and global features with larger receptive fields, making it a promising approach for medical image segmentation.