Scholarlink

Purpose <p>To evaluate the diagnostic performance of an integrated model using intratumoral habitat imaging and peritumoral CT radiomics for preoperative noninvasive prediction of Glypican-3 (GPC3) expression in hepatocellular carcinoma (HCC).​​.</p> Methods <p>A retrospective analysis was performed on preoperative contrast-enhanced CT images and corresponding GPC3 immunohistochemical expression data from 203 patients with pathologically confirmed HCC. Intratumoral habitat features and peritumoral radiomics features (defined within 5&#xa0;mm and 8&#xa0;mm expansion regions from the tumor boundary) were extracted from the CT images. A nomogram was constructed by integrating the habitat Risk score, peritumoral radiomics Rad-score, and selected clinical indicators (including Edmondson grade and microvascular invasion). The diagnostic performance of these radiomics signatures was rigorously assessed through multiple analytical approaches, including discrimination accuracy measured by the area under the receiver operating characteristic curve (AUC) with statistical comparison using DeLong’s test, calibration accuracy evaluated via Hosmer-Lemeshow testing, and clinical utility determined by decision curve analysis across relevant probability thresholds.</p> Results <p>The combined GPC3-RadNomogram model demonstrated significantly superior predictive performance compared to other models in both training and validation cohorts. The AUC values were 0.912 (95% CI: 0.866–0.958) and 0.927 (95% CI: 0.861–0.993) for the training and validation sets, respectively. Hosmer-Lemeshow tests yielded p-values &gt; 0.05 in both cohorts. Decision curve analysis confirmed significant net clinical benefit across clinically reasonable threshold probabilities (15% − 60%). DeLong’s test revealed that habitat features provided significantly higher discriminative power for GPC3 expression than clinical models and peritumoral radiomics models in both cohorts (<i>P</i> &lt; 0.001, |z|&gt;1.96), displaying improved calibration and clinical practicality.</p> Conclusions <p>The CT radiomics model based on habitat analysis enables improved prediction of GPC3 expression in HCC by integrating heterogeneity quantification of intratumoral habitats, peritumoral microenvironment features, and clinicopathological indicators.</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

Noninvasive prediction of Glypican-3 expression in hepatocellular carcinoma using Habitat-based and peritumoral CT radiomics: a nomogram approach

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月26日

查看全文
  • Jiaqi Zhang,
  • Xiaoshu Zhu,
  • Jiamei Qiu,
  • Houhui Shi,
  • Yuting Liu,
  • Jiake Hua,
  • Xushuang Qin,
  • Shanni Dong,
  • Yang Liu,
  • Cuiyun Wu,
  • Jun Chen

摘要

Purpose

To evaluate the diagnostic performance of an integrated model using intratumoral habitat imaging and peritumoral CT radiomics for preoperative noninvasive prediction of Glypican-3 (GPC3) expression in hepatocellular carcinoma (HCC).​​.

Methods

A retrospective analysis was performed on preoperative contrast-enhanced CT images and corresponding GPC3 immunohistochemical expression data from 203 patients with pathologically confirmed HCC. Intratumoral habitat features and peritumoral radiomics features (defined within 5 mm and 8 mm expansion regions from the tumor boundary) were extracted from the CT images. A nomogram was constructed by integrating the habitat Risk score, peritumoral radiomics Rad-score, and selected clinical indicators (including Edmondson grade and microvascular invasion). The diagnostic performance of these radiomics signatures was rigorously assessed through multiple analytical approaches, including discrimination accuracy measured by the area under the receiver operating characteristic curve (AUC) with statistical comparison using DeLong’s test, calibration accuracy evaluated via Hosmer-Lemeshow testing, and clinical utility determined by decision curve analysis across relevant probability thresholds.

Results

The combined GPC3-RadNomogram model demonstrated significantly superior predictive performance compared to other models in both training and validation cohorts. The AUC values were 0.912 (95% CI: 0.866–0.958) and 0.927 (95% CI: 0.861–0.993) for the training and validation sets, respectively. Hosmer-Lemeshow tests yielded p-values > 0.05 in both cohorts. Decision curve analysis confirmed significant net clinical benefit across clinically reasonable threshold probabilities (15% − 60%). DeLong’s test revealed that habitat features provided significantly higher discriminative power for GPC3 expression than clinical models and peritumoral radiomics models in both cohorts (P < 0.001, |z|>1.96), displaying improved calibration and clinical practicality.

Conclusions

The CT radiomics model based on habitat analysis enables improved prediction of GPC3 expression in HCC by integrating heterogeneity quantification of intratumoral habitats, peritumoral microenvironment features, and clinicopathological indicators.