Scholarlink

<p>Alzheimer’s disease (AD) is a chronic neurodegenerative disorder that significantly contributes to the global burden of dementia, affecting over 55&#xa0;million individuals worldwide. Early detection of AD is crucial for timely intervention and effective management of symptoms. Traditional diagnostic methods, such as cognitive tests, MRI, and PET scans, are often costly, invasive, and not readily accessible, especially in low- and middle-income countries. Machine learning (ML) and deep learning (DL) methods have great potential by taking advantage of big clinical and imaging data to achieve accurate and non-invasive early diagnosis. This paper presents a new XRD-SCFNet model, integrating the merits of DenseNet and residual networks along with Spatial Context Fusion (SCF) blocks, for AD stage classification based on MRI images. The proposed XRD-SCFNet model integrates dense connections to preserve low-level features, residual connections to enable deep hierarchical feature learning, and SCF blocks to bridge the gap between local and global features. Bayesian optimization is employed for hyperparameter tuning, and conformal prediction is used to provide confidence intervals for model predictions, enhancing interpretability and reliability. The XRD-SCFNet model achieved an overall accuracy of 96.5% and demonstrated robust performance across all stages of AD, thus providing a very accurate and efficient solution for Alzheimer’s disease stage classification based on MRI images.</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

XRDNet: a Novel Explainable Residual Dense Fusion Network for Alzheimer’s Disease Recognition from MRI Images

  • 期刊论文

  • 发表日期：

    2025年12月10日

查看全文
  • Muhammad John Abbas,
  • Muhammad Attique Khan,
  • Amir Hussain,
  • Sarra Ayouni,
  • Mohamed Maddeh,
  • Fatimah Alhayan

摘要

Alzheimer’s disease (AD) is a chronic neurodegenerative disorder that significantly contributes to the global burden of dementia, affecting over 55 million individuals worldwide. Early detection of AD is crucial for timely intervention and effective management of symptoms. Traditional diagnostic methods, such as cognitive tests, MRI, and PET scans, are often costly, invasive, and not readily accessible, especially in low- and middle-income countries. Machine learning (ML) and deep learning (DL) methods have great potential by taking advantage of big clinical and imaging data to achieve accurate and non-invasive early diagnosis. This paper presents a new XRD-SCFNet model, integrating the merits of DenseNet and residual networks along with Spatial Context Fusion (SCF) blocks, for AD stage classification based on MRI images. The proposed XRD-SCFNet model integrates dense connections to preserve low-level features, residual connections to enable deep hierarchical feature learning, and SCF blocks to bridge the gap between local and global features. Bayesian optimization is employed for hyperparameter tuning, and conformal prediction is used to provide confidence intervals for model predictions, enhancing interpretability and reliability. The XRD-SCFNet model achieved an overall accuracy of 96.5% and demonstrated robust performance across all stages of AD, thus providing a very accurate and efficient solution for Alzheimer’s disease stage classification based on MRI images.