Scholarlink

<p>Monkeypox is a zoonotic disease caused by the monkeypox virus, which is easily transmitted among individuals and poses a significant threat to public health. In this paper, we propose DI-SESCA-MSNET, an enhanced multi-scale neural network for end-to-end monkeypox virus skin lesion classification. Building upon the MSMP-Net framework, our method integrates the Dual-Input Squeeze-and-Excitation Spatial-Channel Attention (DI-SESCA) module, which leverages both channel-wise and spatial attention signals while fusing multi-scale feature maps to strengthen feature representation. We adopt ConvNeXt as the backbone network, incorporating inverse bottleneck layers and large convolution kernels to further boost feature extraction capabilities. Additionally, a multi-scale feature fusion strategy merges the deeper feature maps from multiple stages, enhancing the network’s capacity to represent monkeypox lesion images. On the MSLD v2.0 dataset, DI-SESCA-MSNET achieves accuracy, precision, recall, and F1-score of 89.38 ± 4.20%, 90.13 ± 3.61%, 89.38 ± 4.20%, and 88.83 ± 4.41%, respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

DI-SESCA-MSNET: dual-input squeeze-and-excitation spatial-channel attention enhanced multi-scale network for monkeypox virus skin lesion classification

  • 期刊论文

  • 发表日期：

    2026年7月15日

查看全文
  • Eryang Huan,
  • Hui Dun,
  • Yang Xian

摘要

Monkeypox is a zoonotic disease caused by the monkeypox virus, which is easily transmitted among individuals and poses a significant threat to public health. In this paper, we propose DI-SESCA-MSNET, an enhanced multi-scale neural network for end-to-end monkeypox virus skin lesion classification. Building upon the MSMP-Net framework, our method integrates the Dual-Input Squeeze-and-Excitation Spatial-Channel Attention (DI-SESCA) module, which leverages both channel-wise and spatial attention signals while fusing multi-scale feature maps to strengthen feature representation. We adopt ConvNeXt as the backbone network, incorporating inverse bottleneck layers and large convolution kernels to further boost feature extraction capabilities. Additionally, a multi-scale feature fusion strategy merges the deeper feature maps from multiple stages, enhancing the network’s capacity to represent monkeypox lesion images. On the MSLD v2.0 dataset, DI-SESCA-MSNET achieves accuracy, precision, recall, and F1-score of 89.38 ± 4.20%, 90.13 ± 3.61%, 89.38 ± 4.20%, and 88.83 ± 4.41%, respectively.