With the continuous progress of medical technology and the improvement of public health awareness, people's requirements for accurate analysis of medical images are becoming more and more stringent. However, most medical images have unclear boundaries, and deep learning networks are prone to “learning inertia” during training, which affects the segmentation effect. To address the above issues, we propose an innovative Lightweight Periodic Modeling Medical Image Segmentation Network Based on Supervision Strategy (LPM-Net). In particular, Deep Adaptive Weight Allocation Strategy (DAWAS) guides the network in the right direction and monitors the network to prevent “learning inertia”. This helps the network to update the weight information as well as the deep learning feature information efficiently. In addition, FAN Conv uses Fourier poles for periodic modeling, which improves network modeling capabilities. In order to improve the network even further for mining hidden modeling information, Periodic Implicit Modeling Module (PIMM) is proposed, which is extended by FANConv. This module has the ability to facilitate deep network analysis and learning of potential feature information along with periodic modeling capability. To verify the superiority of the method proposed in this paper, we conducted thorough experimental verifications on three datasets. Our code is available at https://github.com/ououyyILPM_Net.