MPUNet: Multi-Scale Feature Enhancement and Pooling Attention Network for Skin Disease Segmentation

Key Points

• The research aims to develop a more accurate network for skin disease segmentation using advanced feature extraction techniques.
• Designed MPUNet incorporating multi-scale feature enhancement and pooling attention modules.
• Applied to skin disease segmentation datasets including ISIC2018, ISIC2017, and PH2.
• Enhanced ability to extract features while suppressing interference from noise.
• MPUNet achieved an average improvement of 1.57 percentage points in mIoU metric on segmentation datasets.