Focusing the View: Enhancing U-Net with Convolutional Block Attention for Superior Medical Image Segmentation

Detecting and segmenting polyps from endoscopic images is a significant challenge in the medical field, aimed at enhancing the early diagnosis rate of potentially cancerous conditions. To address the limitations of current methods in identifying the complex features of polyps, we have developed an improved version of U-Net, integrating the Convolutional Block Attention Module (CBAM). This combination improves the feature extraction, focuses on critical aspects of polyps, and minimizes unnecessary noise and errors. Testing on multiple datasets has shown that our model significantly outperforms traditional U-Net versions, particularly in detecting small and deformed polyps, while also delivering high computational efficiency suitable for real-world medical applications. This study not only opens new avenues in medical imaging segmentation technology but also has the potential to improve diagnostic and treatment procedures significantly.