Scholarlink

<p>Early detection of emerging infectious diseases is essential for timely public health interventions, yet traditional surveillance systems often miss early, non-specific signals. This study introduces Early Recognition using Neural Information Encoding (ERNIE), an autonomous, disease-agnostic natural language processing framework that applies unsupervised language modelling to unstructured primary care data to detect atypical disease clusters without predefined diagnostic labels. Trained on northern Netherlands data and validated using external datasets, ERNIE identified early COVID-19-like clusters before the first confirmed case, detected RSV-indicative patterns during the 2021 surge, and flagged simulated West Nile Virus cases, while remaining stable during controls (consultation-level recall 0.97; cluster precision 0.82; cluster recall 0.90). By analysing routine clinical text, ERNIE enables early and interpretable detection of emerging health threats without reliance on diagnostic codes or laboratory confirmation. Beyond this, the framework holds potential for identifying atypical clinical patterns in other domains, supporting scalable, AI-driven monitoring across health systems.</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

AI-driven early infectious disease detection in Dutch primary care using BERT and ERNIE

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月23日

查看全文
  • Maarten Homburg,
  • Gijs Danoe,
  • Marjolein Y. Berger,
  • Tim olde Hartman,
  • Jean Muris,
  • Andreas Voss,
  • Axel Hamprecht,
  • Maarten F. Brilman,
  • Lilian L. Peters,
  • Matthijs S. Berends

摘要

Early detection of emerging infectious diseases is essential for timely public health interventions, yet traditional surveillance systems often miss early, non-specific signals. This study introduces Early Recognition using Neural Information Encoding (ERNIE), an autonomous, disease-agnostic natural language processing framework that applies unsupervised language modelling to unstructured primary care data to detect atypical disease clusters without predefined diagnostic labels. Trained on northern Netherlands data and validated using external datasets, ERNIE identified early COVID-19-like clusters before the first confirmed case, detected RSV-indicative patterns during the 2021 surge, and flagged simulated West Nile Virus cases, while remaining stable during controls (consultation-level recall 0.97; cluster precision 0.82; cluster recall 0.90). By analysing routine clinical text, ERNIE enables early and interpretable detection of emerging health threats without reliance on diagnostic codes or laboratory confirmation. Beyond this, the framework holds potential for identifying atypical clinical patterns in other domains, supporting scalable, AI-driven monitoring across health systems.