Scholarlink

<p>Alzheimer’s disease (AD) is a chronic, incurable brain disorder, and early detection is essential for effective management. Traditional detection methods often rely on large, pre-labeled image datasets, which are costly and difficult to compile. To address this, we propose an innovative active learning framework that improves model performance using fewer labeled samples. Conventional active learning techniques often use static selection strategies that lack adaptability. To address this, the method combines deep reinforcement learning (DRL) with a scope loss function (SLF) to improve flexibility. This allows a dynamic balance between exploiting known data and exploring new data opportunities. To reduce hyperparameter sensitivity in DRL, we apply an advanced differential evolution (DE) algorithm. The model was evaluated on the OASIS and ADNI datasets, achieving F-measures of 92.044% and 93.685%, showing its superiority in early AD detection.</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

Automated Alzheimer’s disease detection using active learning model with reinforcement learning and scope loss function

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月8日

查看全文
  • Zhisen He,
  • Vijay Govindarajan,
  • Jing Yang,
  • Mahmoud Abdel-Salam,
  • Zaffar Ahmed Shaikh,
  • Lip Yee Por,
  • Roohallah Alizadehsani,
  • Ru-San Tan,
  • U. Rajendra Acharya

摘要

Alzheimer’s disease (AD) is a chronic, incurable brain disorder, and early detection is essential for effective management. Traditional detection methods often rely on large, pre-labeled image datasets, which are costly and difficult to compile. To address this, we propose an innovative active learning framework that improves model performance using fewer labeled samples. Conventional active learning techniques often use static selection strategies that lack adaptability. To address this, the method combines deep reinforcement learning (DRL) with a scope loss function (SLF) to improve flexibility. This allows a dynamic balance between exploiting known data and exploring new data opportunities. To reduce hyperparameter sensitivity in DRL, we apply an advanced differential evolution (DE) algorithm. The model was evaluated on the OASIS and ADNI datasets, achieving F-measures of 92.044% and 93.685%, showing its superiority in early AD detection.