To address the problem of low segmentation accuracy of multifracture muscle lesion sites in MRI images of patients with lumbar disc herniation, this paper proposes a new model to improve the U2net network with the goal that the encoding and decoding subnetworks are interconnected by a series of nested jump paths, and this redesigned jump connection is mainly used to reduce the semantic missing of feature maps in the encoding and decoding subnetworks. For this purpose, the jump connections in the middle of RSU-7, RSU-6,RSU-5,RSU-4 in the U2Net model are redesigned, and the RSU-4F part remains unchanged. In addition, the channel attention module was added to extract high quality multifractal muscle features, which effectively helps information flow in the network by using an attention mechanism to increase expressiveness, focus on important features and suppress unnecessary features. To verify the effectiveness of the model, experiments are conducted on the multi-cleft muscle MRI image dataset, and it is found that MIOU, DICE, and HD are improved compared with U-Net, U2-Net, and U-Net++ network structures. The experimental results show that the algorithm proposed in this paper has a good effect on MRI image segmentation of multifidus muscle, which can assist doctors to make judgments on the condition.