Medical image segmentation is critical for computer-aided medical diagnosis, both convolutional neural networks and the Transformer model have received widespread attention. However, convolutional Neural Networks and Transformers are each limited by global information and computational complexity. Recent studies have shown that the state space model (SSM) represented by Vision Mamba is able to maintain linear computational complexity and to simulate long-range interactions. Inspired by this, we propose the High-order Morphological Method Vision Mamba (HMM-VMamba) in this paper to further enhance the focus on lesion regions. Among its components, the proposed High-order Morphological Method Selective-Scan-2D (HMM-SS2D) exploits high-order information channel interactions to explore morphological knowledge. Moreover, the combination of Local Selective-Scan-2D, External Multi-scale Attention Mechanism, and Convolutional Neural Network improves the capability of local boundary feature learning. We evaluated comparison and ablation experiments on three open-source medical image datasets, and the results demonstrate that HMM-VMamba exhibits strong learning capabilities in medical image segmentation. Github