Scholarlink

This study proposes Dual-InfNet, a novel deep-learning architecture for automated lung infection segmentation in chest CT scans to aid COVID-19 diagnosis. Dual-InfNet addresses challenges like varying infection characteristics and low contrast by integrating a Convolutional Block Attention Module (CBAM) within each encoder layer for enhanced feature refinement. CBAM generates attention maps that adaptively refine features, while a parallel partial decoder aggregates high-level features for global infection localization. Finally, the model utilizes implicit reverse attention and explicit edge attention to refine boundaries and enhance representations. Extensive experiments demonstrate that Dual-InfNet outperforms existing models, paving the way for advanced AI applications in healthcare.

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

Enhanced Chest CT Infection Segmentation with Reverse Attention and CBAM

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Nomaiya Bashree,
  • Tareque Bashar Ovi,
  • Hussain Nyeem,
  • Md Abdul Wahed

摘要

This study proposes Dual-InfNet, a novel deep-learning architecture for automated lung infection segmentation in chest CT scans to aid COVID-19 diagnosis. Dual-InfNet addresses challenges like varying infection characteristics and low contrast by integrating a Convolutional Block Attention Module (CBAM) within each encoder layer for enhanced feature refinement. CBAM generates attention maps that adaptively refine features, while a parallel partial decoder aggregates high-level features for global infection localization. Finally, the model utilizes implicit reverse attention and explicit edge attention to refine boundaries and enhance representations. Extensive experiments demonstrate that Dual-InfNet outperforms existing models, paving the way for advanced AI applications in healthcare.