<p>Medical image segmentation is a key technology in computer-aided diagnosis systems. However, the accuracy of the existing segmentation models often falls short in practical applications. To improve the accuracy of the model, this paper proposes a novel U-shaped network called MMP-Net, meticulously crafted for medical image segmentation tasks. The MMP-Net incorporates three core modules: the multi-kernel convolution module, which enhances the multi-receptive field representation of the model; the multi-layer-guided channel attention module, which combines the features from different encoder layers, and the combined features are used to guide the channel attention; the phase-guided Laplacian convolution module, which leverages the boundary sensitivity of Laplacian convolution kernels to effectively capture edge gradient changes and local detail textures in images. The proposed MMP-Net has been validated with three metrics (Dice, HD95, and IOU) across five public medical image datasets (ISIC2017, ISIC2018, BUSI, COVID-19, PH2). Experimental results demonstrate that the MMP-Net outperforms other popular models in all these three metrics with moderate model parameters (2.03M) and computation (5.36G FLOPs). This achievement offers an efficient and accurate solution for medical image segmentation tasks, making it particularly suitable for mobile healthcare and edge computing scenarios. The source code will be available at <a href="https://github.com/liyiwei-png/MMP-Net.git">https://github.com/liyiwei-png/MMP-Net.git</a></p>