Scholarlink

Background <p>Incidental gallbladder cancer (IGBC) is often diagnosed only during or after cholecystectomy, and preoperative identification remains challenging. This study preliminarily explored factors contributing to IGBC missed diagnosis and attempted to develop an ultrasound radiomics‑based identification model.</p> Methods <p>A retrospective cohort of 62 IGBC and 78 non‑incidental GBC (NIGBC) patients who were consecutively enrolled between 2016/01–2025/12 was analyzed. Clinical, laboratory, imaging, pathological, and immunohistochemical features were compared. From ultrasound images, 1220 radiomics features were extracted; after stability (ICC &gt; 0.75), redundancy removal (Spearman |ρ|&gt; 0.90), and LASSO regression, nine features were retained. Seven machine learning algorithms were used to exploratorily build radiomics‑only models. Potential clinical predictors were identified by logistic regression, and a clinical‑only and a combined model were attempted. Performance was preliminarily evaluated using area under the curve (AUC).</p> Results <p>IGBC showed higher gallstone prevalence (91.9% vs. 53.8%, <i>P</i> &lt; 0.001) and a “benign masquerade” laboratory profile (higher albumin (ALB), high-density lipoprotein cholesterol (HDL‑C); lower ratio of albumin to globulin (RAR); fewer elevated CA19‑9). Pathologically, IGBC was predominantly infiltrative (77.4% vs. 28.2%, <i>P</i> &lt; 0.001), had earlier T stage compared to NIGBC (27.4% vs. 12.8%, <i>P</i> = 0.007), and exhibited lower Ki‑67 high expression (67.7% vs. 83.3%, <i>P</i> = 0.031) and weaker Topoisomerase II-alpha (Topo II) staining (<i>P</i> = 0.005). The extreme gradient boosting (XGB) radiomics model achieved a validation AUC of 0.865 (accuracy 0.833), suggesting potential discriminative ability. Gallstones (OR = 9.484) and growth pattern (OR = 0.230) might be independent clinical predictors. The clinical‑only model had AUC 0.830, and the combined model AUC 0.860, with no significant benefit over radiomics‑only.</p> Conclusion <p>Preoperative missed diagnosis of IGBC may be associated with gallstone‑related inflammation, infiltrative growth, early T stage, seemingly normal laboratory findings, and low proliferation marker expression. Although conventional ultrasound hardly identifies IGBC, its tumor heterogeneity might be quantified by radiomics. The XGB model showed preliminary ability to distinguish IGBC from NIGBC and holds potential as a non‑invasive tool for preoperative risk stratification, but findings require validation in larger multicenter cohorts.</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

Factors contributing to the missed diagnosis of incidental gallbladder cancer (IGBC): an integrated analysis of clinical features and ultrasound radiomics

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月10日

查看全文
  • Zhiping Pan,
  • Xu Sun,
  • Zhenhua Shen,
  • Feng Cen,
  • Qiang Yan

摘要

Background

Incidental gallbladder cancer (IGBC) is often diagnosed only during or after cholecystectomy, and preoperative identification remains challenging. This study preliminarily explored factors contributing to IGBC missed diagnosis and attempted to develop an ultrasound radiomics‑based identification model.

Methods

A retrospective cohort of 62 IGBC and 78 non‑incidental GBC (NIGBC) patients who were consecutively enrolled between 2016/01–2025/12 was analyzed. Clinical, laboratory, imaging, pathological, and immunohistochemical features were compared. From ultrasound images, 1220 radiomics features were extracted; after stability (ICC > 0.75), redundancy removal (Spearman |ρ|> 0.90), and LASSO regression, nine features were retained. Seven machine learning algorithms were used to exploratorily build radiomics‑only models. Potential clinical predictors were identified by logistic regression, and a clinical‑only and a combined model were attempted. Performance was preliminarily evaluated using area under the curve (AUC).

Results

IGBC showed higher gallstone prevalence (91.9% vs. 53.8%, P < 0.001) and a “benign masquerade” laboratory profile (higher albumin (ALB), high-density lipoprotein cholesterol (HDL‑C); lower ratio of albumin to globulin (RAR); fewer elevated CA19‑9). Pathologically, IGBC was predominantly infiltrative (77.4% vs. 28.2%, P < 0.001), had earlier T stage compared to NIGBC (27.4% vs. 12.8%, P = 0.007), and exhibited lower Ki‑67 high expression (67.7% vs. 83.3%, P = 0.031) and weaker Topoisomerase II-alpha (Topo II) staining (P = 0.005). The extreme gradient boosting (XGB) radiomics model achieved a validation AUC of 0.865 (accuracy 0.833), suggesting potential discriminative ability. Gallstones (OR = 9.484) and growth pattern (OR = 0.230) might be independent clinical predictors. The clinical‑only model had AUC 0.830, and the combined model AUC 0.860, with no significant benefit over radiomics‑only.

Conclusion

Preoperative missed diagnosis of IGBC may be associated with gallstone‑related inflammation, infiltrative growth, early T stage, seemingly normal laboratory findings, and low proliferation marker expression. Although conventional ultrasound hardly identifies IGBC, its tumor heterogeneity might be quantified by radiomics. The XGB model showed preliminary ability to distinguish IGBC from NIGBC and holds potential as a non‑invasive tool for preoperative risk stratification, but findings require validation in larger multicenter cohorts.