Scholarlink

Alzheimer’s disease is a progressive neurodegenerative condition that impairs cognitive function and quality of life among the elderly. This study introduces EnXAI Net, an ensemble deep learning framework that integrates SAINT, FT Transformer, LSTM, and DaNet architectures. Leveraging clinical and genetic data from the ADNI dataset, the model classifies subjects into cognitively normal, mild cognitive impairment and Alzheimer’s disease groups using a weighted voting strategy. Key features include age, gender, MMSE, ADAS13, and APOE genotype. To enhance interpretability, explainable AI methods such as SHAP and LIME are employed to quantify feature contributions. Experimental results demonstrate that the model achieves a classification accuracy exceeding 97%, while offering transparency in decision making. The proposed approach highlights the potential of modern AI technologies in clinical diagnostics and contributes to the advancement of personalized healthcare within the context of digital transformation. Inference averages 4.76 s per subject, enabling near-real-time clinical decision support workflows.

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

EnXAI-Net: Enhancing Diagnostic Accuracy in Alzheimer’s Disease via Explainable Ensemble Learning Model

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Khac-Tuong Nguyen,
  • Anh- Cang Phan

摘要

Alzheimer’s disease is a progressive neurodegenerative condition that impairs cognitive function and quality of life among the elderly. This study introduces EnXAI Net, an ensemble deep learning framework that integrates SAINT, FT Transformer, LSTM, and DaNet architectures. Leveraging clinical and genetic data from the ADNI dataset, the model classifies subjects into cognitively normal, mild cognitive impairment and Alzheimer’s disease groups using a weighted voting strategy. Key features include age, gender, MMSE, ADAS13, and APOE genotype. To enhance interpretability, explainable AI methods such as SHAP and LIME are employed to quantify feature contributions. Experimental results demonstrate that the model achieves a classification accuracy exceeding 97%, while offering transparency in decision making. The proposed approach highlights the potential of modern AI technologies in clinical diagnostics and contributes to the advancement of personalized healthcare within the context of digital transformation. Inference averages 4.76 s per subject, enabling near-real-time clinical decision support workflows.