Scholarlink

<p>The grading assessment of Alzheimer's disease based on neuropsychology mainly relies on observing the degree of daily life impairment of patients, completing the Mini Mental State Scale, Global Deterioration Scale, etc., and obtaining clinical diagnostic conclusions based on score levels. This approach is inevitably influenced by subjective factors of the observer, and the scale developed based on specific hypothetical models may not fully conform to the true statistical patterns, resulting in misdiagnosis. This study designs a federated learning algorithm model with 8 clients based on brain magnetic resonance images of individuals of different severity levels involved in the disease on a considerable scale. Each client has independent image data resources, and the neural network of the client is synchronously trained by a central server to aggregate model parameters in real time. The sharing of medical data resources is completed while fully protecting data privacy. After 380 iterations, the model's evaluation accuracy on the test set can reach over 98%. This method integrates several data islands into a larger pool of medical imaging resources, and trains and fits an intelligent clinical evaluation model based on a special method to effectively increase the data scale, providing an ideal solution for objective grade diagnosis of Alzheimer's disease based on brain imaging.</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

Intelligent model for Alzheimer's disease imaging assessment based on federated learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月28日

查看全文
  • Jia Yao

摘要

The grading assessment of Alzheimer's disease based on neuropsychology mainly relies on observing the degree of daily life impairment of patients, completing the Mini Mental State Scale, Global Deterioration Scale, etc., and obtaining clinical diagnostic conclusions based on score levels. This approach is inevitably influenced by subjective factors of the observer, and the scale developed based on specific hypothetical models may not fully conform to the true statistical patterns, resulting in misdiagnosis. This study designs a federated learning algorithm model with 8 clients based on brain magnetic resonance images of individuals of different severity levels involved in the disease on a considerable scale. Each client has independent image data resources, and the neural network of the client is synchronously trained by a central server to aggregate model parameters in real time. The sharing of medical data resources is completed while fully protecting data privacy. After 380 iterations, the model's evaluation accuracy on the test set can reach over 98%. This method integrates several data islands into a larger pool of medical imaging resources, and trains and fits an intelligent clinical evaluation model based on a special method to effectively increase the data scale, providing an ideal solution for objective grade diagnosis of Alzheimer's disease based on brain imaging.