Scholarlink

<p>Enhanced CT-based radiomics features and clinical indicators were integrated to develop a risk prediction model for intrahepatic recurrence (IR) in early-stage hepatocellular carcinoma (HCC) patients undergoing initial radiofrequency ablation (RFA). This multicenter retrospective study enrolled 261 patients with HCC treated with RFA. Patients were grouped into training, internal validation, and external validation cohorts. Six machine learning algorithms were used to build model for predicting IR from preoperative arterial-phase CT images. The model with the highest area under the receiver operating characteristic curve (AUC) was selected as the optimal radiomics model (Radscore). A clinical model was constructed using multivariate Cox regression, and a combined clinical-radiomics nomogram was developed by integrating the Radscore and significant clinical indicators. IR occurred in 121 patients. KNN-Radscore demonstrated superior predictive performance. Multivariate Cox analysis identified maximum tumor diameter (hazard ratio [HR] 1.04, 95% confidence interval [CI]: 1.02–1.07), tumor number (HR 5.33, 95% CI: 2.24–12.68), and tumor margin (HR 3.00, 95% CI: 1.72–5.22) as independent predictors. The clinical model achieved C-indexes of 0.722 (training), 0.712 (internal validation) and 0.766 (external validation), while the combined model yielded 0.733, 0.726 and 0.787, respectively. The combined model’s AUCs for 1‑ and 2‑year survival were 0.760 and 0.762 in training, 0.685 and 0.785 in internal validation, and 0.857 and 0.856 in external validation. The combined clinical‑radiomics model, integrating a KNN‑based radiomics signature with three independent clinical predictors (maximum tumor diameter, tumor number, and irregular margin), exhibited robust performance for preoperatively predicting intrahepatic recurrence after initial RFA in early‑stage HCC, thus providing a reference for individualized risk stratification.</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 CT-based radiomics nomogram for predicting intrahepatic recurrence of HCC after radiofrequency ablation

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月28日

查看全文
  • Lin Chen,
  • Mengchen Yuan,
  • Meng Wang,
  • Yuhan Zhou,
  • Qingbo Huang,
  • Qi Yang,
  • Zhigang Zhou,
  • Xiuqin Jia

摘要

Enhanced CT-based radiomics features and clinical indicators were integrated to develop a risk prediction model for intrahepatic recurrence (IR) in early-stage hepatocellular carcinoma (HCC) patients undergoing initial radiofrequency ablation (RFA). This multicenter retrospective study enrolled 261 patients with HCC treated with RFA. Patients were grouped into training, internal validation, and external validation cohorts. Six machine learning algorithms were used to build model for predicting IR from preoperative arterial-phase CT images. The model with the highest area under the receiver operating characteristic curve (AUC) was selected as the optimal radiomics model (Radscore). A clinical model was constructed using multivariate Cox regression, and a combined clinical-radiomics nomogram was developed by integrating the Radscore and significant clinical indicators. IR occurred in 121 patients. KNN-Radscore demonstrated superior predictive performance. Multivariate Cox analysis identified maximum tumor diameter (hazard ratio [HR] 1.04, 95% confidence interval [CI]: 1.02–1.07), tumor number (HR 5.33, 95% CI: 2.24–12.68), and tumor margin (HR 3.00, 95% CI: 1.72–5.22) as independent predictors. The clinical model achieved C-indexes of 0.722 (training), 0.712 (internal validation) and 0.766 (external validation), while the combined model yielded 0.733, 0.726 and 0.787, respectively. The combined model’s AUCs for 1‑ and 2‑year survival were 0.760 and 0.762 in training, 0.685 and 0.785 in internal validation, and 0.857 and 0.856 in external validation. The combined clinical‑radiomics model, integrating a KNN‑based radiomics signature with three independent clinical predictors (maximum tumor diameter, tumor number, and irregular margin), exhibited robust performance for preoperatively predicting intrahepatic recurrence after initial RFA in early‑stage HCC, thus providing a reference for individualized risk stratification.