解码动脉瘤血栓:EVAR术后瘤囊缩小的影像组学特征
单中心回顾性研究纳入94例择期EVAR的肾下腹主动脉瘤患者,从术前CTA提取血栓影像组学特征,评估其与12个月随访瘤囊缩小的关联。29例出现瘤囊缩小,两个影像组学特征与缩小显著相关,临床-影像组学联合模型AUC为0.76。
要点
- 术前血栓影像组学特征与EVAR术后瘤囊缩小相关,联合临床变量可提供互补信息。
- 联合模型预测瘤囊缩小的AUC为0.76、敏感度0.70、特异度0.75、准确率0.73。
- 结果需在更大队列中验证后方可用于临床决策。
核心结论
- 术前血栓影像组学特征与EVAR术后瘤囊缩小相关,与临床变量整合后提供互补信息。
- 研究支持血栓影像组学用于刻画EVAR术后瘤囊重塑的潜在价值,但需在更大队列中验证。
研究设计与样本
- 单中心回顾性研究,纳入94例因肾下腹主动脉瘤行择期EVAR的患者。
- 按12个月随访时瘤囊是否缩小进行分层,分析术前动脉期CTA。
- 使用3D Slicer手动分割瘤囊并以腔内血栓(ILT)为感兴趣区。
- 通过PyRadiomics结合高斯拉普拉斯滤波提取530个影像组学特征,经标准化后以Wilcoxon-Mann-Whitney检验、Pearson相关过滤及基于AIC的特征选择进行筛选。
- 比较临床-影像组学联合模型、纯影像组学模型与纯临床模型的预测性能。
主要发现
- 94例患者中29例出现瘤囊缩小。
- 两个影像组学特征与瘤囊缩小显著相关。
- 较年轻年龄(P = 0.02)、较大动脉瘤直径(P = 0.02)及均匀血栓质地与瘤囊缩小相关。
- 无慢性阻塞性肺疾病(COPD)与瘤囊未缩小相关。
- 其他评估的临床或解剖变量与瘤囊缩小均无显著关联。
关键数值与推荐
- 联合临床-影像组学模型性能最佳:AUC 0.76、敏感度0.70、特异度0.75、准确率0.73。
- 较年轻年龄P = 0.02、较大动脉瘤直径P = 0.02与瘤囊缩小相关。
- 共提取530个影像组学特征,采用高斯拉普拉斯滤波。
临床怎么用
- 对拟行EVAR的肾下腹主动脉瘤患者,术前CTA血栓影像组学特征可辅助预判瘤囊缩小模式。
- 联合临床变量(年龄、瘤径、COPD状态)构建的模型可支持术后随访风险分层。
- 目前尚处于研究阶段,需外部大样本验证后再考虑纳入临床决策。
局限性
- 单中心回顾性设计,样本量94例、缩小事件29例,结论外推受限。
- 影像组学特征与模型未经外部队列验证。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND: Aneurysm sac shrinkage after endovascular aneurysm repair (EVAR) is a surrogate marker of effective aneurysm exclusion and favorable outcomes. Radiomics enables quantitative imaging analysis beyond visual assessment and is increasingly applied in cardiovascular imaging. Analysis of intraluminal thrombus (ILT) may provide additional information for identifying patients with different patterns of sac remodeling after EVAR. METHODS: This single-center retrospective study assessed whether radiomic features of the ILT extracted from preoperative computed tomography angiography (CTA), alone and in combination with clinical and anatomical variables, were associated with aneurysm sac shrinkage after EVAR. Patients undergoing elective EVAR for infrarenal abdominal aortic aneurysm (AAA) were included and stratified according to sac shrinkage at a 12-month follow up. Preoperative arterial-phase CTA scans were analyzed. The aneurysm sac was manually segmented using three-dimensional (3D) slicer, and the ILT was selected as the region of interest. A total of 530 radiomic features were extracted using PyRadiomics with Laplacian of Gaussian filtering. Features were standardized and evaluated using the Wilcoxon-Mann-Whitney test, Pearson's correlation filtering, and Akaike Information Criterion (AIC)-based feature selection. Predictive performance was compared among combined clinical-radiomic, radiomic-only, and clinical-only models. RESULTS: Among 94 patients, 29 showed aneurysm sac shrinkage. Two radiomic features were significantly associated with shrinkage. Younger age (P = 0.02), larger aneurysm diameter (P = 0.02), and homogeneous thrombus texture were associated with shrinkage, while the absence of chronic obstructive pulmonary disease (COPD) was associated with lack of shrinkage. None of the other evaluated clinical or anatomical variables showed a significant association with sac shrinkage. The combined clinical-radiomic model achieved the best performance (area under the curve [AUC] 0.76, sensitivity 0.70, specificity 0.75, and accuracy 0.73). CONCLUSION: Preoperative thrombus radiomic features were associated with subsequent sac shrinkage and provided complementary information when integrated with clinical variables. These findings support the potential role of thrombus radiomics in the characterization of sac remodeling after EVAR and warrant validation in larger cohorts.
引用
- Rascio L, Boldrini L, Boccia E, et al. Decoding the Aneurysm Thrombus: Radiomic Signatures of Sac Shrinkage after Endovascular Aortic Repair. Ann Vasc Surg. 2026. PMID: 42830122. doi:10.1016/j.avsg.2026.09.025
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。