Scholarlink

<p>Lung disease remains a leading global health challenge, necessitating accurate and automated diagnostic systems to assist clinicians. In this study, we propose a novel framework for lung disease classification that integrates deep convolutional feature learning with Graph Attention Networks (GATs). As the initial step, rich spatial representations are extracted from chest X-ray images by a CNN backbone. These characteristics are now structured into graph nodes, where anatomical regions or patch-based embeddings are connected as per the spatial and semantic relationships. GAT module learns the significance of neighbouring nodes in an adaptive way with the help of the attention mechanism. This mechanism allows the model to learn local and global dependencies in the lung. This system enhances discriminative capabilities of the features and hence it is useful in identifying complex pathologies. Studies on standard benchmark datasets on the chest X-ray show that the proposed method will significantly outperform state-of-the-art CNN-based and graph convolutional baselines on the grounds of the accuracy, AUC, and F1-score. This shows the effective use of deep features learning and attention-controlled graph modelling to achieve accurate, reliable and better lung disease classification.</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

A hybrid deep learning approach for lung disease classification using deep CNNs and graph attention networks

  • 期刊论文

  • 发表日期：

    2026年4月24日

查看全文
  • Sandhya Devi,
  • Pallavi Khatri

摘要

Lung disease remains a leading global health challenge, necessitating accurate and automated diagnostic systems to assist clinicians. In this study, we propose a novel framework for lung disease classification that integrates deep convolutional feature learning with Graph Attention Networks (GATs). As the initial step, rich spatial representations are extracted from chest X-ray images by a CNN backbone. These characteristics are now structured into graph nodes, where anatomical regions or patch-based embeddings are connected as per the spatial and semantic relationships. GAT module learns the significance of neighbouring nodes in an adaptive way with the help of the attention mechanism. This mechanism allows the model to learn local and global dependencies in the lung. This system enhances discriminative capabilities of the features and hence it is useful in identifying complex pathologies. Studies on standard benchmark datasets on the chest X-ray show that the proposed method will significantly outperform state-of-the-art CNN-based and graph convolutional baselines on the grounds of the accuracy, AUC, and F1-score. This shows the effective use of deep features learning and attention-controlled graph modelling to achieve accurate, reliable and better lung disease classification.