Scholarlink

<p>Medical imaging has become a foundation of modern healthcare, providing critical insights for the diagnosis and management of a wide range of diseases, yet accurate interpretation remains challenging due to indistinct lesion boundaries and complex anatomical variations. This work proposed an advanced dual-path deep learning framework tailored for the classification of medical images, with the primary goal of enhancing diagnostic accuracy and reliability. The proposed model innovatively combines edge and texture features through a dual-path architecture that integrates learned edge extraction with partial ResNet50 layers for texture representation. A novel light dual-gated feature interaction module (LightDGFIM) cross-attention mechanism effectively fuses low-level structural information with high-level semantic features, improving sensitivity to anatomical boundaries and lesion margins, which is a key challenge in medical image analysis. The model’s robustness and generalization are validated on multiple organ-specific datasets, including Alzheimer, Eye Cataract, and Pneumonia, using comprehensive evaluation metrics such as accuracy, precision, recall, F1-score, and AUC. The results demonstrate consistent improvements over state-of-the-art pretrained models, highlighting the clinical potential of edge-aware feature fusion for enhanced AI-assisted diagnosis and decision-making. An extensive ablation study further validates the contribution of each model component, demonstrating the importance of edge-aware feature fusion and the dual attention mechanism.</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

Edge-aware dual path network for medical image classification

  • 期刊论文

  • 发表日期：

    2026年2月5日

查看全文
  • Muhammad Hamza Mehmood,
  • Xinming Zhang,
  • Emmanuel U. Ugwu

摘要

Medical imaging has become a foundation of modern healthcare, providing critical insights for the diagnosis and management of a wide range of diseases, yet accurate interpretation remains challenging due to indistinct lesion boundaries and complex anatomical variations. This work proposed an advanced dual-path deep learning framework tailored for the classification of medical images, with the primary goal of enhancing diagnostic accuracy and reliability. The proposed model innovatively combines edge and texture features through a dual-path architecture that integrates learned edge extraction with partial ResNet50 layers for texture representation. A novel light dual-gated feature interaction module (LightDGFIM) cross-attention mechanism effectively fuses low-level structural information with high-level semantic features, improving sensitivity to anatomical boundaries and lesion margins, which is a key challenge in medical image analysis. The model’s robustness and generalization are validated on multiple organ-specific datasets, including Alzheimer, Eye Cataract, and Pneumonia, using comprehensive evaluation metrics such as accuracy, precision, recall, F1-score, and AUC. The results demonstrate consistent improvements over state-of-the-art pretrained models, highlighting the clinical potential of edge-aware feature fusion for enhanced AI-assisted diagnosis and decision-making. An extensive ablation study further validates the contribution of each model component, demonstrating the importance of edge-aware feature fusion and the dual attention mechanism.