Scholarlink

This study involved the development and evaluation of a novel deep neural network model for Alzheimer’s disease and related dementias (ADRD) phenotyping. The model was initially trained on a large cohort of 100,000 cases and controls and subsequently fine-tuned using a smaller, expert-reviewed dataset of 1,200 individuals. The final fine-tuned model achieved an Area Under the Receiver Operating Characteristic Curve of 0.832. For further validation, the model’s predictive capability was assessed in a separate randomly selected patient cohort comprising individuals without an ADRD diagnosis from 2009 to 2018. The survival analysis shows that patients with higher predicted ADRD risk scores exhibited a significantly increased incidence of developing ADRD after their index date within five years.

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

Improving Alzheimer’s Disease and Related Dementias Phenotyping in Electronic Health Records Through Transfer Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yijun Shao,
  • Ying Yin,
  • Debby Tsuang,
  • Phillip Ma,
  • Edward Zamrini,
  • Ali Ahmed,
  • Charles Faselis,
  • Katherine Wilson,
  • Karl Brown,
  • Qing Zeng-Treitler

摘要

This study involved the development and evaluation of a novel deep neural network model for Alzheimer’s disease and related dementias (ADRD) phenotyping. The model was initially trained on a large cohort of 100,000 cases and controls and subsequently fine-tuned using a smaller, expert-reviewed dataset of 1,200 individuals. The final fine-tuned model achieved an Area Under the Receiver Operating Characteristic Curve of 0.832. For further validation, the model’s predictive capability was assessed in a separate randomly selected patient cohort comprising individuals without an ADRD diagnosis from 2009 to 2018. The survival analysis shows that patients with higher predicted ADRD risk scores exhibited a significantly increased incidence of developing ADRD after their index date within five years.