Edge Enhancement Based Semi-supervised Medical Image Segmentation Method

Annotating 3D medical images is a time-consuming and labor-intensive process, which has led to the development of semi-supervised learning (SSL) approaches to alleviate the reliance on extensive labeled data. However, weak edge issue in medical images remains a major challenge, limiting the performance of existing SSL methods. To address this issue, we propose Edge Enhancement V-Net (EEVNet), a semi-supervised segmentation network designed to enhance edge features. Specifically, we integrate the Simple Attention Module (SimAM) into the encoder and propose an Edge Enhancement Module (EEM) to refine edge predictions and improve segmentation performance on weak edges. Additionally, a dual-network co-training strategy is employed to enhance pseudo-label reliability. Experiments on the AMOS dataset demonstrate that our method performs better than some state-of-the-art methods.