IVIM MRI预测成人弥漫性胶质瘤IDH突变状态的系统综述与Meta分析
该系统综述纳入13项研究,评估IVIM参数D、D*和f预测成人弥漫性胶质瘤IDH突变状态的诊断准确性。D参数汇总敏感度0.808、特异度0.762、AUC 0.849,表现最一致;D*和f准确性较低且异质性大。证据等级为低或极低,阈值多为事后选定,限制了临床转化。
要点
- D参数是IVIM中预测IDH突变状态最一致的指标,AUC 0.849
- 各研究阈值均为事后选定,阅片时不宜直接套用具体截断值
- IVIM差异不能视为IDH特异性,因肿瘤类型和分级与分子状态存在混杂
核心结论
- D参数在IVIM各参数中显示最一致的诊断准确性,汇总AUC为0.849(95%CI 0.804-0.884)
- D*(AUC 0.727)和f(AUC 0.764)准确性较低且异质性更大,f的结果方向不一致
- 证据确定性低(D)至极低(D*和f),且阈值多为事后选定,限制了临床转化应用
研究设计与样本
- 系统综述与诊断试验准确性Meta分析,检索5个数据库截至2026年5月13日
- 共纳入13项研究,9项研究贡献了至少一项Meta分析
- 采用QUADAS-2评估偏倚风险,双变量随机效应模型合并数据,GRADE评估证据确定性
- D参数合并分析纳入7项研究、499例患者
关键影像表现/征象
- D参数反映组织扩散系数,IDH突变型与野生型胶质瘤之间D值存在差异
- D*为假扩散系数,f为灌注分数,两者在IDH突变预测中表现不稳定
- f参数在不同研究中方向不一致,部分研究显示突变型f值更高,部分相反
诊断标准与关键数值
- D参数汇总敏感度0.808(95%CI 0.735-0.864),特异度0.762(95%CI 0.696-0.817)
- D参数汇总AUC为0.849(95%CI 0.804-0.884),基于7项研究、499例患者
- D*汇总AUC为0.727(7项研究),f汇总AUC为0.764(6项研究)
- 全部13项研究Index Test偏倚风险均为高风险,因阈值均为事后选定
- b值方案、拟合方法、ROI定义和阈值在各研究间差异显著
临床应用与阅片要点
- 阅片时可将D参数作为IVIM中相对可靠的参考指标,但需结合常规序列和分子病理综合判断
- D*和f的临床应用价值不确定,不宜单独用于IDH突变状态判断
- 因各研究阈值不同且为事后选定,报告IVIM定量值时宜注明所采用的b值方案和后处理方式
- 解释IVIM差异时需注意肿瘤类型和级别与IDH状态存在混杂,差异不等于分子特异性
局限性
- 全部纳入研究的Index Test偏倚风险为高,因诊断阈值均为事后选定
- 各研究间b值方案、拟合方法、ROI定义和阈值差异显著,存在明显异质性
- D参数证据确定性为低,D*和f为极低
- 肿瘤类型和分级与分子状态存在混杂,IVIM差异不能视为IDH特异
- 缺乏标准化和前瞻性外部验证
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)。
英文原文(摘要)· 点击展开
OBJECTIVES: To estimate the diagnostic accuracy of intravoxel incoherent motion (IVIM) MRI parameters, including diffusion coefficient (D), pseudo-diffusion coefficient (D*), and perfusion fraction (f), in predicting isocitrate dehydrogenase (IDH) mutation status in adult diffuse glioma, and identify methodological factors underlying heterogeneity across the IVIM literature.
MATERIALS AND METHODS: Five databases were searched through 13 May 2026. Two reviewers independently screened studies, extracted data, and assessed risk of bias using QUADAS-2. Per-study 2 × 2 tables were reconstructed where possible and pooled using a bivariate random-effects (Reitsma) model. Certainty of evidence was assessed using GRADE.
RESULTS: Thirteen studies were included; nine contributed to at least one meta-analysis. For D (7 studies, 499 patients), pooled sensitivity was 0.808 (95% confidence interval [CI] 0.735-0.864), specificity 0.762 (95% CI 0.696-0.817), and summary area under the curve (AUC) 0.849 (95% CI 0.804-0.884). All 13 studies had high Index Test risk of bias because thresholds were selected post hoc. B-value schemes, fitting methods, ROI definitions, and thresholds also varied substantially. D* (7 studies, AUC 0.727) and f (6 studies, AUC 0.764) were less accurate and more heterogeneous. Findings for f were directionally inconsistent. Certainty was low for D and very low for D* and f.
CONCLUSION: D showed the most consistent accuracy among IVIM parameters. However, IVIM differences cannot be considered IDH-specific because tumour type and grade were confounded with molecular status. Low-certainty evidence and post hoc thresholds limit clinical translation. Evidence for D* and f remains limited. Standardization and prospective external validation are needed before clinical implementation.
引用
- Khan, Stasso, Melkus, et al. Intravoxel incoherent motion (IVIM) MRI for predicting IDH mutation status in adult diffuse glioma: a systematic review and diagnostic test accuracy meta-analysis. Eur J Radiol. 2026. PMID: 42759286. doi:10.1016/j.ejrad.2026.113253
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。