Scholarlink

<p>Coronavirus Disease 2019 (COVID-19) is a rapidly spreading acute respiratory illness with severe pulmonary sequelae. Therefore, the accurate diagnosis of various lung diseases is crucial for effective clinical treatment. This study introduces a novel chest X-ray lung disease classification method named EFTransNet. We utilize EfficientNet as the backbone network to extract low-level feature maps and incorporate the Coordinate Attention Augmentation Module to enhance spatial feature perception. To address the limitations of traditional convolutional neural networks in capturing long-range dependencies, we propose the Adaptive Feature Integration and Enhancement module, which uses a Transformer encoder for self-attention encoding. Additionally, we integrate the Vision Transformer structure, and through an adaptive weighted attention gate mechanism, low-level features and high-level semantic features are effectively integrated. To address the class imbalance in lung disease datasets, we design the Focal Loss with Class Attention, assigning higher weights to difficult-to-classify samples, thereby improving the accuracy of minority classes. We evaluated EFTransNet on publicly available chest X-ray and CT scan datasets. Experimental results show that the accuracy for Virus reached 0.8754, and the accuracy for Bacteria reached 0.8925. Compared to baseline methods, the performance of minority classes significantly improved, validating the effectiveness of our approach.</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

FETransNet: an enhanced lung disease classification approach combining EfficientNet and transformer with adaptive focal loss

  • 期刊论文

  • 发表日期：

    2025年8月9日

查看全文
  • Zhentao Hu,
  • Hongyu Chen,
  • Weiqiang Mei,
  • Yihan Wang

摘要

Coronavirus Disease 2019 (COVID-19) is a rapidly spreading acute respiratory illness with severe pulmonary sequelae. Therefore, the accurate diagnosis of various lung diseases is crucial for effective clinical treatment. This study introduces a novel chest X-ray lung disease classification method named EFTransNet. We utilize EfficientNet as the backbone network to extract low-level feature maps and incorporate the Coordinate Attention Augmentation Module to enhance spatial feature perception. To address the limitations of traditional convolutional neural networks in capturing long-range dependencies, we propose the Adaptive Feature Integration and Enhancement module, which uses a Transformer encoder for self-attention encoding. Additionally, we integrate the Vision Transformer structure, and through an adaptive weighted attention gate mechanism, low-level features and high-level semantic features are effectively integrated. To address the class imbalance in lung disease datasets, we design the Focal Loss with Class Attention, assigning higher weights to difficult-to-classify samples, thereby improving the accuracy of minority classes. We evaluated EFTransNet on publicly available chest X-ray and CT scan datasets. Experimental results show that the accuracy for Virus reached 0.8754, and the accuracy for Bacteria reached 0.8925. Compared to baseline methods, the performance of minority classes significantly improved, validating the effectiveness of our approach.