Abstract:

Objective To compare the accuracy of radiologically estimated liver volume by computedtomography（CT）and magnetic resonance imaging（MRI）. Methods Cochrane library，abstract databases 〔Medline，Embase，China biology medicine disc（CMBdisc）〕and full text databases（Springer，OVID， Sciencedirect，Ebsco，CNKI，WanFang and VIP）were searched for published papers on whole liver volumetrywith CT or MRI modality. Inclusion criteria were based on validity criteria for diagnostic research published by theCochrane collaboration. Eligible papers were grouped according to imaging modality（CT or MRI）and evaluated forquality before data extraction. With meta-analysis package Meta of R software，the statistical model used to pool the

result was selected according to result of heterogeneity test. Group-wise meta-analysis and comparison of included studies of CT and MRI liver volumetry were performed. The pooled result was weighted mean difference between estimated and true liver volume. Finally sensitivity and publish bias analysis was performed. Results 17 articles with 37 studies including 351 patients meet the inclusion criteria. There was no heterogeneity among included studiesof each group（CT ：I2 ＝ 0%，Q ＝ 4.93，P ＝ 0.999 8 ；MRI ：I2 ＝ 0%，Q ＝ 0.92，P ＝ 1）. Fixed effect modelwas selected for the meta-analysis and the weighted mean difference was 21.706 3 ml〔95％ confidence interval （95%CI）＝ -8.015 4 - 51.464 0 ml）〕for the CT group and 20.951 8 ml（95％ CI ＝ -60.436 0 - 102.339 5 ml）for the MRI group. There was no statistically significant difference between the true and radiologically estimated whole liver volume of each group（CT ：Z ＝ 1.429 7，P ＝ 0.152 8 ；MRI ：Z ＝ 0.504 6，P ＝ 0.613 9）and between pooled results of each group（Q ＝ 0，P ＝ 1）. Sensitivity analysis revealed the pooled result of each group was robust which was not affected by excluding any of the included studies or changing the effect model. Funnel plot and publish bias analysis revealed no publish bias（CT ：t ＝ -1.882 0，P ＝ 0.075 2 ；MRI ：t ＝ 0.001 5，P ＝ 0.998 9）. Conclusion Accurate liver volumetry could be performed with CT or MRI images. The imaging acquisitionmethods should be selected according to the patients' general status with respect to the patients' benefits.

摘要：

目的 比较采用体层摄影术（CT）、磁共振成像（MRI）图像进行肝体积影像学测量的准确性。方法 电子检索 Cochrane 图书馆、中英文文摘数据库（Medline、Embase、中国生物医学文摘数据库）和期刊全文数据库（Springer、OVID、Sciencedirect、Ebsco、CNKI、万方、维普等），获得利用 CT 或 MRI 设备测量全肝体积的中英文文献，根据 Cochrane 协作网推荐的诊断试验纳入标准筛选文献，根据 CT、MRI 成像设备分组后进行质量评价，采用基于统计分析软件 R 的循证医学软件包 Meta 对纳入文献提取信息进行异质性检验，根据异质性检验结果选择恰当的效应量合并模型进行荟萃分析，分别对纳入的 CT 和 MRI 研究进行加权合并及组间比较，合并效应量为影像学测量肝体积和实际肝体积的加权平均差，最后进行敏感性分析和发表性偏倚分析。结果 符合纳入标准的 17 篇文献报道了 37 项研究，包括 351 例患者。异质性检验显示 CT 组（I2 ＝ 0%，Q ＝ 4.93，P ＝ 0.999 8）和 MRI 组（I2 ＝ 0%，Q ＝ 0.92，P ＝ 1）纳入研究均不具异质性。各组效应量合并采用固定效应模型，CT 和 MRI 组影像学测量肝体积和实际肝体积的加权均数差值分别为 21.706 3 ml〔95％可信区间（95%CI）＝ -8.015 4 ～ 51.464 0 ml〕和 20.951 8 ml（95％ CI ＝ -60.436 0 ～102.339 5 ml）。两组的影像学测量肝体积和实际肝体积差异均无统计学意义（CT 组：Z ＝ 1.429 7，P ＝ 0.152 8； MRI 组 ：Z ＝ 0.504 6，P ＝ 0.613 9），两组的汇总结果差异也无统计学意义（Q ＝ 0，P ＝ 1）。各组汇总结果稳定性好，敏感性分析中逐一排除纳入研究或改变效应量合并模型时，合并结果无明显变化，各纳入研究不存在发表性偏倚，漏斗图各纳入研究对称分布，Deeks 线性回归分析不存在显著性（CT 组：t ＝ -1.882 0， P ＝ 0.075 2 ；MRI 组 ：t ＝ 0.001 5，P ＝ 0.998 9）。结论 采用 CT 和 MRI 图像都可以进行准确的肝体积影像学测量，可以根据患者的具体情况，选择适合患者利益的图像采集方法。