电子科技2026，Vol.39Issue(4)：1-7,7.DOI:10.16180/j.cnki.issn1007-7820.2026.04.001
基于多尺度通道注意力的特征融合皮肤病分割方法
Multi-Scale Channel Attention Based on Feature Fusion Method for Skin Disease Segmentation
摘要
Abstract
In view of the problem of low image segmentation accuracy caused by factors such as irregular shapes,blurred edges,and foreign body occlusion of melanoma,this study proposes a MSFFNet(Multi Scale Feature Fusion Network).In the encoding stage,a pyramid split attention module is introduced to expand the receptive field and capture feature information at different scales.The channel attention mechanism is used to recalibrate the weights of different channels,and then point-by-point multiplication fusion is performed with the original feature space.The Dice Loss function is adopted for end-to-end optimization to alleviate the negative impact of class imbalance in sam-ples and further improve the segmentation performance of the network model.The segmentation performance of the model is verified on the ISIC2018 skin disease image dataset.Experimental results show that the Dice and IoU(Inter-section over Union)of the proposed method are 89.40%and 82.27%respectively.Compared with mainstream seg-mentation algorithms,the proposed segmentation method is more similar to the results of manual segmentation by doc-tors.关键词
皮肤病图像/图像分割/多尺度/注意力机制/特征融合/金字塔注意力/通道注意力/损失函数Key words
skin disease images/image segmentation/multi-scale/attention mechanisms/feature fusion/pyramid attention/channel attention/loss function分类
信息技术与安全科学引用本文复制引用
徐坤财,邓湛,刘璇,张宁,卢家东..基于多尺度通道注意力的特征融合皮肤病分割方法[J].电子科技,2026,39(4):1-7,7.基金项目
贵州省青年科技人才成长项目(黔教技[2024]279)Guizhou Province Youth Science and Technology Talent Growth Project(Qian Jiao Ji[2024]279) （黔教技[2024]279）
