Scholarlink

<p>Sjögren’s disease (SjD) is a common systemic autoimmune disease that remains difficult to diagnose early due to non-specific symptoms and lack of definitive biomarkers. This multicentre retrospective cohort study analysed data from 34,958 patients across three hospitals in China to assess the diagnostic potential of routine laboratory tests and to develop a robust, low-cost artificial intelligence model, the Sjögren Multi-criterion Feature Integration Framework (SMFIF). This study is registered with ClinicalTrials.gov (NCT06982482). The model was built using 16 optimal features selected through ensemble learning and SHAP analysis, and validated internally (<i>n</i> = 9329) and externally (<i>n</i> = 545). SMFIF achieved high diagnostic performance, with AUCs of 0.929, 0.934, and 0.964 in testing, internal, and external validation sets, respectively, outperforming conventional biomarkers such as anti-SSA/Ro and ANA. Calibration curves and confusion matrices confirmed its reliability. SMFIF is publicly available and provides probabilities of SjD based on laboratory data, offering a practical diagnostic tool for clinical use.</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

A multi-criterion feature integration framework for accurate diagnosis of Sjögren’s disease using routine laboratory tests

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月27日

查看全文
  • Shu Liu,
  • Guanhao Wu,
  • Mengge Pan,
  • Qi Sun,
  • Chenjing Gao,
  • Xianming Long,
  • Chaogui Tang,
  • Xinran Yuan,
  • Lingyun Sun

摘要

Sjögren’s disease (SjD) is a common systemic autoimmune disease that remains difficult to diagnose early due to non-specific symptoms and lack of definitive biomarkers. This multicentre retrospective cohort study analysed data from 34,958 patients across three hospitals in China to assess the diagnostic potential of routine laboratory tests and to develop a robust, low-cost artificial intelligence model, the Sjögren Multi-criterion Feature Integration Framework (SMFIF). This study is registered with ClinicalTrials.gov (NCT06982482). The model was built using 16 optimal features selected through ensemble learning and SHAP analysis, and validated internally (n = 9329) and externally (n = 545). SMFIF achieved high diagnostic performance, with AUCs of 0.929, 0.934, and 0.964 in testing, internal, and external validation sets, respectively, outperforming conventional biomarkers such as anti-SSA/Ro and ANA. Calibration curves and confusion matrices confirmed its reliability. SMFIF is publicly available and provides probabilities of SjD based on laboratory data, offering a practical diagnostic tool for clinical use.