Scholarlink

<p>Preoperative differentiation between xanthogranulomatous cholecystitis (XGC) and gallbladder cancer (GBC) remains challenging due to overlapping clinical and imaging features. This multicenter retrospective study developed a machine learning (ML) model, LIDGAX, using preoperative clinical, imaging, and laboratory data from 1246 patients (554 XGC, 692 GBC). Twelve variables were identified as independent predictors via multivariate logistic regression and least absolute shrinkage and selection operator analyses. LIDGAX achieved area under the curve (AUC) values of 0.94 (internal validation) and 0.88 (external testing), outperforming the other five ML models. Calibration and decision curve analyses demonstrated its superior clinical utility. Compared to six radiologists, LIDGAX improved sensitivity (1.2–8.5%), specificity (0.0–4.6%), and balanced accuracy (1.8–6.6%), while reducing average diagnostic time per patient by 30.44–35.76 s. LIDGAX was deployed on an open-source online platform, maintaining high performance (AUC 0.95, accuracy 0.92). This non-invasive tool shows strong potential for clinical translation in preoperative differentiation of XGC and GBC.</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

Machine learning model for differentiating xanthogranulomatous cholecystitis and gallbladder cancer in multicenter largescale study

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月1日

查看全文
  • Ke Zhang,
  • Jiajia He,
  • Weiyue Ji,
  • Qunyan Pan,
  • Weilv Xiong,
  • Liping Wang,
  • Weiqi Sun,
  • Liting Xie,
  • Tianan Jiang

摘要

Preoperative differentiation between xanthogranulomatous cholecystitis (XGC) and gallbladder cancer (GBC) remains challenging due to overlapping clinical and imaging features. This multicenter retrospective study developed a machine learning (ML) model, LIDGAX, using preoperative clinical, imaging, and laboratory data from 1246 patients (554 XGC, 692 GBC). Twelve variables were identified as independent predictors via multivariate logistic regression and least absolute shrinkage and selection operator analyses. LIDGAX achieved area under the curve (AUC) values of 0.94 (internal validation) and 0.88 (external testing), outperforming the other five ML models. Calibration and decision curve analyses demonstrated its superior clinical utility. Compared to six radiologists, LIDGAX improved sensitivity (1.2–8.5%), specificity (0.0–4.6%), and balanced accuracy (1.8–6.6%), while reducing average diagnostic time per patient by 30.44–35.76 s. LIDGAX was deployed on an open-source online platform, maintaining high performance (AUC 0.95, accuracy 0.92). This non-invasive tool shows strong potential for clinical translation in preoperative differentiation of XGC and GBC.