PVT-BE: A Boundary-aware Polyp Segmentation-based Model

Recently, deep learning-based methods for automated polyp segmentation have been limited by factors such as low contrast between polyps and surrounding tissues and problems associated with detection devices. These limitations result in the inability to accurately distinguish polyps from the surrounding mucosal tissues and identify polyps effectively. To address these issues, this paper proposes the PVT-BE model. Specifically, we first employ the Pyramid Vision Transformer encoder to learn multi-scale feature representations. We propose a simple and effective Edge Enhancement Module (EEM) for enhancing polyp features and filtering image noise. To maximize the capture of polyp edge information, we introduce the boundary-aware attention (BAAM) module, and our PVT-BE model is quantitatively and qualitatively evaluated on five challenging datasets. The best performance was achieved on four datasets and ranked second on one dataset. Demonstrating its strong potential for clinical application in polyp segmentation. Ablation studies further validate the contribution of the EEM and BAAM modules.