nVR模型：预测原发性经动脉化疗栓塞术后肝细胞癌的LI-RADS治疗反应及不可行复发情况。
The nVR model: prediction of LI-RADS treatment response nonviable recurrence of hepatocellular carcinoma after primary transarterial chemoembolization.
作者信息
Zhang Shuhang, Wang Weilang, Zhang Xiuming, Zhong Binyan, Feng Feng, Cai Wu, Li Binrong, Menon Varsha Ajith, Zhou Shuwei, Zhang Teng, Chen Xunjun, Ju Shenghong, Wang Yuan-Cheng
机构信息
Department of Radiology, Nurturing Center of Jiangsu Province for State Laboratory of AI Imaging & Interventional Radiology, Zhongda Hospital, Southeast University, Nanjing, China.
Department of Radiology, Jiangsu Cancer Hospital, Nanjing, China.
出版信息
Abdom Radiol (NY). 2025 Apr 11. doi: 10.1007/s00261-025-04924-w.
OBJECTIVES
This study aims to build a prediction model based on MRI features and clinical characteristics for early tumor recurrence (< 12 months) in lesions which evaluated as LR-TR nonviable at first follow up (1-2 months) after transaterial chemoembolization (TACE).
METHODS
This multicenter retrospective study included consecutive patients with hepatocellular carcinoma (HCC) who underwent initial TACE from five centers between February 2015 and October 2022. Additionally, patients from a completed clinical trial (NCT03113955) were also included for this retrospective analysis. Nineteen baseline imaging features and six first follow-up imaging features were evaluated for lesions classified as LR-TR nonviable at the first post-TACE imaging assessment. A predictive model integrating clinical and MRI features for early recurrence of LR-TR nonviable lesions was developed with logistic regression analyses and validated on a 5-fold cross-validation. The model's performance was further validated by analyzing relapse-free survival (RFS) using Kaplan-Meier curves.
RESULTS
Non-smooth margin (OR: 4.69; 95% CI: 1.91, 11.56; P = 0.001) and peritumoral hyperintensity on T2-weighted imaging or diffusion-weighted imaging (OR: 6.12; 95% CI: 2.53, 14.80; P < 0.001) were identified as independent risk factors for early recurrence of LR-TR nonviable lesions. The non-Viable Recurrence (nVR) model, developed using these two features, achieved area under the receiver operating characteristic curve of 0.759 (0.734, 0.784) in the training cohort and 0.765 (0.632, 0.898) in the validation cohort. Patients were stratified into high-risk and low-risk groups based on the Youden index, with RFS significantly differing between these groups in both the training (P < 0.001) and validation cohorts (P = 0.009).
CONCLUSION
Based on two radiologic features, the nVR model has demonstrated relatively reliable efficacy in predicting early post-TACE recurrence.
目的
本研究旨在基于MRI特征和临床特征构建一个预测模型，用于预测经动脉化疗栓塞术（TACE）后首次随访（1 - 2个月）时评估为LR-TR不可存活的病变的早期肿瘤复发（<12个月）。
方法
这项多中心回顾性研究纳入了2015年2月至2022年10月期间在五个中心接受初次TACE的连续肝细胞癌（HCC）患者。此外，来自一项已完成临床试验（NCT03113955）的患者也被纳入此次回顾性分析。对TACE后首次成像评估中分类为LR-TR不可存活的病变，评估了19个基线影像特征和6个首次随访影像特征。通过逻辑回归分析建立了一个整合临床和MRI特征的LR-TR不可存活病变早期复发预测模型，并在5折交叉验证中进行了验证。使用Kaplan-Meier曲线分析无复发生存期（RFS）进一步验证了该模型的性能。
结果
边缘不光滑（OR：4.69；95%CI：1.91，11.56；P = 0.001）以及T2加权成像或扩散加权成像上的瘤周高信号（OR：6.12；95%CI：2.53，14.80；P < 0.001）被确定为LR-TR不可存活病变早期复发的独立危险因素。使用这两个特征建立的非存活复发（nVR）模型在训练队列中的受试者操作特征曲线下面积为0.759（0.734，0.784），在验证队列中为0.765（0.63, 0.898）。根据约登指数将患者分为高风险和低风险组，在训练队列（P < 0.001）和验证队列（P = 0.009）中，两组的RFS均有显著差异。
结论
基于两个放射学特征，nVR模型在预测TACE后早期复发方面显示出相对可靠的疗效。
Zotero 插件