基于MRI影像组学评估肝细胞癌微血管侵犯的Meta分析

目的 通过Meta分析系统评价基于MRI影像组学术前评估肝细胞癌（HCC）微血管侵犯（MVI）的诊断效能。 方法 检索PubMed、Cochrane图书馆、EMBASE、Web of Science、中国知网和万方数据库,检索时间为自建库以来至2021年3月公开发表的有关MRI影像组学评估HCC MVI的临床研究的中英文文献。根据纳入和排除标准对文献进行筛选并提取数据。根据诊断准确性研究质量评价工具-2（QUADAS-2）评价纳入文献的质量,采用RevMan5.4评价偏倚风险及临床适用性。采用Stata15.1软件计算纳入研究的合并敏感度、合并特异度、合并阳性似然比、合并阴性似然比及合并诊断比值比。绘制森林图和总受试者操作特征（SROC）曲线,并计算曲线下面积（AUC）。采用不一致指数（I²）和Chochrane Q指数评估异质性,并采用阈值效应和敏感性分析探讨异质性来源。采用Stata15.1软件绘制漏斗图评估是否存在发表偏倚。 结果 共纳入10项研究,包括817例病人,纳入研究的合并敏感度、特异度、阳性似然比、阴性似然比、诊断比值比及AUC值分别为0.82（95%CI：0.76~0.87）、0.84（95%CI：0.77~0.89）、5.1（95%CI：3.4~7.7）、0.22（95%CI：0.16~0.29）、24（95%CI：13~44）和0.89（95%CI：0.86~0.91）。敏感度的I²为27.37%,特异度的I²为62.19%,纳入研究间异质性较大。Deek’s漏斗图显示不存在发表偏倚（P=0.83）。 结论 基于MRI影像组学术前预测HCC 的MVI有较高的准确性,可作为术前预测HCC MVI的一种无创性方法。

Using radiomic features to assess microvascular invasion in hepatocellular carcinoma based on MRI imaging:A Meta-analysis

Objective To systematically assess the diagnostic performance of radiomics based on MRI for microvascular invasion (MVI) in hepatocellular carcinoma (HCC) before surgery by Meta-analysis. Method The databases of PubMed, Cochrane library, EMBASE, Web of Science, CNKI and Wanfang were searched for literature about predicting MVI in HCC by MRI radiomics published from their inception to March 2021. The literature was screened according to inclusion and exclusion criteria, and data was extracted. The qualities of the included studies were assessed according to quality assessment of diagnostic accuracy studies-2(QUADAS-2). Rev Man 5.4 was performed to evaluate the risk of bias and clinical applicability. Stata 15.1 software was used to calculated pooled sensitivity, specificity, positive likelihood ratio, negative likelihood ratio and diagnostic odds ratio. Forest plots and summary receiver operating characteristic (SROC) curve were plotted and the area under the curve (AUC) was calculated. The heterogeneity between studies was evaluated with I² and Chochrane Q. Threshold effect and sensitivity analysis were used to explore the source of heterogeneity. Stata15.1 was used to draw funnel plots to assess whether there was publication bias. Result Ten studies involving 817 patients were included. The pooled sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, diagnostic odds ratio and AUC were 0.82 (95%CI: 0.76-0.87), 0.84(95%CI: 0.77-0.89), 5.1(95%CI: 3.4-7.7), 0.22(95%CI: 0.16-0.29), 24(95%CI: 13-44) and 0.89(95%CI: 0.86-0.91), respectively. The I² of the sensitivity was 27.37%, and the I² of the specificity was 62.19%, which showed large heterogeneity existed among the included studies. Deek’s test indicated there was no publication bias (P=0.83). Conclusion MRI radiomics has high accuracy for predicting MVI in HCC, and it can be served as a noninvasive method for assessing MVI in HCC before surgery.