Scholarlink

<p>Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that significantly impacts cognitive function and quality of life. The early diagnosis and an accurate severity assessment of AD are always the point of target to the medical fraternity. To fulfill the dual objective this study presents a stacked ensemble deep learning framework for the automated classification and severity ranking of Alzheimer’s disease using MRI scans. The framework integrates multiple deep learning models including EfficientNet-B7, Xception, and Inception-ResNet-V2 to extract rich spatial features from MRI scans and is subsequently integrated using a rank-based fusion approach. The proposed method is evaluated on large-scale MRI OASIS and ADNI datasets of Alzheimer’s patients. According to experimental results, the suggested ensemble performs well on the OASIS dataset, achieving 97.8% accuracy, 96.5% sensitivity, and 98.2% specificity. The model outperforms the current single-model baselines with an accuracy of 98.1%, sensitivity of 97.3%, and specificity of 98.5% on the ADNI dataset. Additionally, the framework is excellent at assessing the severity of Alzheimer’s disease, making it a trustworthy tool for clinical decision support.</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 stacked ensemble deep learning framework for Alzheimer’s severity ranking and classification using MRI scans

  • 期刊论文

  • 发表日期：

    2025年7月25日

查看全文
  • Nidhi Pandey,
  • Oshin Sharma

摘要

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that significantly impacts cognitive function and quality of life. The early diagnosis and an accurate severity assessment of AD are always the point of target to the medical fraternity. To fulfill the dual objective this study presents a stacked ensemble deep learning framework for the automated classification and severity ranking of Alzheimer’s disease using MRI scans. The framework integrates multiple deep learning models including EfficientNet-B7, Xception, and Inception-ResNet-V2 to extract rich spatial features from MRI scans and is subsequently integrated using a rank-based fusion approach. The proposed method is evaluated on large-scale MRI OASIS and ADNI datasets of Alzheimer’s patients. According to experimental results, the suggested ensemble performs well on the OASIS dataset, achieving 97.8% accuracy, 96.5% sensitivity, and 98.2% specificity. The model outperforms the current single-model baselines with an accuracy of 98.1%, sensitivity of 97.3%, and specificity of 98.5% on the ADNI dataset. Additionally, the framework is excellent at assessing the severity of Alzheimer’s disease, making it a trustworthy tool for clinical decision support.