Scholarlink

This paper proposes a multi-stage hybrid model that combines a 3D self-attention mechanism with a convolutional neural network for the classification of Alzheimer’s disease. First, the model uses a 3D multi-head self-attention mechanism to globally process MRI images and identify global correlations between brain MRIs. Then, local information is identified through convolutional layers. Secondly, we use the AAL template to divide the MRI images into four regions: the hippocampus, amygdala, entorhinal cortex, and the whole brain, and give higher weights to high-correlation regions such as the hippocampus, thereby enhancing feature learning and improving classification accuracy by fusing local and global information.

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

Fusion of Convolutional Neural Networks Based on Self-attention Mechanism for Alzheimer’s Disease Image Classification

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Kuo Zhang,
  • Lei Xiao

摘要

This paper proposes a multi-stage hybrid model that combines a 3D self-attention mechanism with a convolutional neural network for the classification of Alzheimer’s disease. First, the model uses a 3D multi-head self-attention mechanism to globally process MRI images and identify global correlations between brain MRIs. Then, local information is identified through convolutional layers. Secondly, we use the AAL template to divide the MRI images into four regions: the hippocampus, amygdala, entorhinal cortex, and the whole brain, and give higher weights to high-correlation regions such as the hippocampus, thereby enhancing feature learning and improving classification accuracy by fusing local and global information.