Scholarlink

Objective <p>This study aims to identify the optimal Ki-67 cutoff for predicting HCC recurrence and to develop a radiomics model based on CT arterial phase features for non-invasive preoperative Ki-67 assessment.</p> Methods <p>A total of 180 HCC patients who underwent radical resection were enrolled. The prognostic value of Ki-67 was analyzed using multivariate Cox proportional hazards regression models, and the optimal Ki-67 expression threshold for recurrence prediction was determined by the area under the ROC curve (AUC).Radiomics features from both intratumoral and peritumoral regions were extracted from preoperative CT arterial-phase images. Key features were selected to develop a radiomics prediction model.The predictive performances of the radiomics, clinical, and combined models were compared, and key influencing features for the optimal model were identified through Shapley additive explanations (SHAP) analysis.</p> Results <p>High Ki-67 expression (≥ 25%) was identified as an independent risk factor for postoperative HCC recurrence (hazard ratio [HR] = 9.923, 95% confidence interval [CI]: 3.383–29.108, <i>P</i> &lt; 0.001). The recurrence-free survival rate was significantly lower in the high Ki-67 expression group compared to the low expression group (<i>P</i> = 0.009). In the testing set, the radiomics model utilizing random forest (RF) algorithm demonstrated superior predictive performance for Ki-67 expression (AUC = 0.869, sensitivity = 0.733, specificity = 0.846). Its performance significantly exceeded that of the clinical model (AUC: 0.869 vs. 0.540, <i>P</i> = 0.005), while the combined model showed no significant improvement compared to the radiomics model alone (AUC: 0.870 vs. 0.869, <i>P</i> = 0.979). SHAP analysis revealed that the peritumoral texture heterogeneity feature (peri_wavelet_LLH_firstorder_Kurtosis) contributed most significantly to the radiomics model (SHAP value = + 0.17).</p> Conclusion <p>High Ki-67 expression (≥ 25%) is a key prognostic factor for postoperative HCC recurrence. The CT arterial phase radiomics model allows non-invasive preoperative Ki-67 assessment, emphasizing the role of peritumoral heterogeneity.</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

Ki-67 expression correlates with hepatocellular carcinoma recurrence and is predictable using radiomics features

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年9月23日

查看全文
  • Haiyang Nong,
  • Yongyi Cen,
  • Zhaolin Pan,
  • Yin Huang,
  • Jiaxin Chen,
  • Deyou Huang,
  • Ke Ding

摘要

Objective

This study aims to identify the optimal Ki-67 cutoff for predicting HCC recurrence and to develop a radiomics model based on CT arterial phase features for non-invasive preoperative Ki-67 assessment.

Methods

A total of 180 HCC patients who underwent radical resection were enrolled. The prognostic value of Ki-67 was analyzed using multivariate Cox proportional hazards regression models, and the optimal Ki-67 expression threshold for recurrence prediction was determined by the area under the ROC curve (AUC).Radiomics features from both intratumoral and peritumoral regions were extracted from preoperative CT arterial-phase images. Key features were selected to develop a radiomics prediction model.The predictive performances of the radiomics, clinical, and combined models were compared, and key influencing features for the optimal model were identified through Shapley additive explanations (SHAP) analysis.

Results

High Ki-67 expression (≥ 25%) was identified as an independent risk factor for postoperative HCC recurrence (hazard ratio [HR] = 9.923, 95% confidence interval [CI]: 3.383–29.108, P < 0.001). The recurrence-free survival rate was significantly lower in the high Ki-67 expression group compared to the low expression group (P = 0.009). In the testing set, the radiomics model utilizing random forest (RF) algorithm demonstrated superior predictive performance for Ki-67 expression (AUC = 0.869, sensitivity = 0.733, specificity = 0.846). Its performance significantly exceeded that of the clinical model (AUC: 0.869 vs. 0.540, P = 0.005), while the combined model showed no significant improvement compared to the radiomics model alone (AUC: 0.870 vs. 0.869, P = 0.979). SHAP analysis revealed that the peritumoral texture heterogeneity feature (peri_wavelet_LLH_firstorder_Kurtosis) contributed most significantly to the radiomics model (SHAP value = + 0.17).

Conclusion

High Ki-67 expression (≥ 25%) is a key prognostic factor for postoperative HCC recurrence. The CT arterial phase radiomics model allows non-invasive preoperative Ki-67 assessment, emphasizing the role of peritumoral heterogeneity.