Scholarlink

<p>Lung cancer remains one of the leading causes of cancer-related mortality worldwide, highlighting the urgent need for early and accurate diagnostic approaches. This study presents a dual-channel channel attention-based hybrid deep learning framework for lung cancer classification with comparatively high accuracy. Two publicly available datasets, IQ-OTH/NCCD (CT scan) and LC25000 (histopathology), were employed to evaluate the proposed model. The framework combines fine-tuned transfer learning from pretrained VGG-16 and EfficientNetB0 backbones with a channel attention mechanism, a gated recurrent unit (GRU), and a customize classification head. Model performance was evaluated using five-fold cross-validation, and explainable artificial intelligence techniques were employed to explain model interpretability. The proposed framework achieved 99.55% of lung cancer classification accuracy on CT scan images and 99.95% on histopathological images, demonstrating competitive and superior performance compared to recent approaches reported in the literature. These findings highlight the potential of proposed hybrid deep learning framework to advance automated lung cancer diagnosis and provide a technical foundation for potential clinically validated decision-making.</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

Interpretable Lung Cancer Detection via Channel Attention Driven Hybrid Deep Learning Model

  • 期刊论文

  • 发表日期：

    2026年6月15日

查看全文
  • Sumaiya Binthe Sazzad,
  • Imtia Islam,
  • Md. Bipul Hossain

摘要

Lung cancer remains one of the leading causes of cancer-related mortality worldwide, highlighting the urgent need for early and accurate diagnostic approaches. This study presents a dual-channel channel attention-based hybrid deep learning framework for lung cancer classification with comparatively high accuracy. Two publicly available datasets, IQ-OTH/NCCD (CT scan) and LC25000 (histopathology), were employed to evaluate the proposed model. The framework combines fine-tuned transfer learning from pretrained VGG-16 and EfficientNetB0 backbones with a channel attention mechanism, a gated recurrent unit (GRU), and a customize classification head. Model performance was evaluated using five-fold cross-validation, and explainable artificial intelligence techniques were employed to explain model interpretability. The proposed framework achieved 99.55% of lung cancer classification accuracy on CT scan images and 99.95% on histopathological images, demonstrating competitive and superior performance compared to recent approaches reported in the literature. These findings highlight the potential of proposed hybrid deep learning framework to advance automated lung cancer diagnosis and provide a technical foundation for potential clinically validated decision-making.