This paper introduces SegMamba, a novel 3D medical image segmentation model that leverages the Mamba architecture, a state space model (SSM) known for its efficiency in modeling long-range dependencies. Unlike traditional convolutional neural networks (CNNs) and transformers, SegMamba excels in capturing global features across the entire volume of 3D medical images while maintaining high computational efficiency. The model incorporates a tri-orientated Mamba (ToM) module to enhance sequential modeling from three directions and a gated spatial convolution (GSC) module to improve spatial feature representation. The authors also introduce a new large-scale dataset, CRC-500, for 3D colorectal cancer segmentation, containing 500 annotated 3D CT scans. Extensive experiments on multiple datasets demonstrate that SegMamba outperforms existing methods in terms of segmentation accuracy and efficiency. The model shows superior performance in capturing long-range dependencies and maintaining high inference speed, making it a promising approach for 3D medical image segmentation tasks. The code and dataset are publicly available for further research.This paper introduces SegMamba, a novel 3D medical image segmentation model that leverages the Mamba architecture, a state space model (SSM) known for its efficiency in modeling long-range dependencies. Unlike traditional convolutional neural networks (CNNs) and transformers, SegMamba excels in capturing global features across the entire volume of 3D medical images while maintaining high computational efficiency. The model incorporates a tri-orientated Mamba (ToM) module to enhance sequential modeling from three directions and a gated spatial convolution (GSC) module to improve spatial feature representation. The authors also introduce a new large-scale dataset, CRC-500, for 3D colorectal cancer segmentation, containing 500 annotated 3D CT scans. Extensive experiments on multiple datasets demonstrate that SegMamba outperforms existing methods in terms of segmentation accuracy and efficiency. The model shows superior performance in capturing long-range dependencies and maintaining high inference speed, making it a promising approach for 3D medical image segmentation tasks. The code and dataset are publicly available for further research.