定量MRI预测肾细胞癌WHO/ISUP分级:系统综述与诊断meta分析
该系统综述与诊断meta分析纳入20项术前MRI评估肾细胞癌WHO/ISUP分级的研究,其中12项可行定量合并。基于ADC的7项研究识别高级别肿瘤的合并敏感度0.84、特异度0.57、AUC 0.71;5项MRI影像组学/深度学习研究合并敏感度0.79、特异度0.86、AUC 0.90。定量MRI尤其扩散指标对分级准确性中等,影像组学/深度学习模型诊断性能较高但研究间异质性大。
要点
- ADC鉴别高低级别RCC敏感度较高(0.84)但特异度有限(0.57),需结合其他序列。
- 低级别RCC的ADC值显著高于高级别,平均差0.21×10⁻³ mm²/s。
- MRI影像组学/深度学习模型合并AUC达0.90,但研究间一致性较低。
核心结论
- 定量MRI尤其扩散衍生指标对预测RCC的WHO/ISUP分级具有中等准确性,影像组学/深度学习模型诊断性能更高但研究间一致性较差,临床推广前需标准化多参数协议、外部验证和决策影响研究。
研究设计与样本
- 系统检索PubMed、Embase、Scopus和Web of Science自建库至2025年8月,纳入经组织病理学证实RCC且行术前MRI、评估定量MRI生物标志物(扩散、弛豫测量、化学交换饱和转移、影像组学)与WHO/ISUP分级关系的研究,共20项,其中12项可行定量meta分析(7项ADC研究、5项MRI影像组学/深度学习研究),采用随机效应模型合并诊断准确性及ADC均数差。
诊断标准与关键数值
- 7项扩散加权MRI研究评估基于ADC的RCC分级,识别高级别疾病的合并敏感度0.84(95%CI 0.77-0.89)、特异度0.57(95%CI 0.51-0.63)、汇总AUC 0.71。
- 低级别肿瘤ADC值显著高于高级别肿瘤,平均差0.21×10⁻³ mm²/s(95%CI 0.11-0.30×10⁻³ mm²/s)。
- 5项MRI影像组学/深度学习研究合并敏感度0.79(95%CI 0.64-0.89)、特异度0.86(95%CI 0.74-0.93)、AUC 0.90。
临床应用与阅片要点
- 定量MRI可辅助无创RCC分级和活检分诊,ADC对排除高级别有一定价值但特异度有限,影像组学/深度学习模型经外部验证后可能改善术前风险分层。
局限性
- 影像组学/深度学习模型研究间一致性较低;临床实施前需标准化多参数协议、外部验证和决策影响研究。
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 synthesise evidence on quantitative MRI biomarkers for predicting World Health Organisation/International Society of Urological Pathology (WHO/ISUP) grade in renal cell carcinoma (RCC).
MATERIALS AND METHODS: We systematically searched PubMed, Embase, Scopus, and Web of Science from inception to August 2025 for patients with histopathologically proven RCC who underwent preoperative MRI. Eligible studies evaluated quantitative MRI biomarkers (diffusion, relaxometry, chemical exchange saturation transfer, radiomics) against WHO/ISUP grade. For diffusion-weighted imaging and radiomics/deep learning, we performed random-effects diagnostic meta-analyses and pooled mean differences in apparent diffusion coefficient (ADC) between low- and high-grade tumours.
RESULTS: Twenty studies were included; quantitative meta-analysis was feasible for 12 (seven ADC studies and five MRI-inclusive radiomics/deep learning studies). Seven diffusion-weighted MRI studies evaluating apparent diffusion coefficient-based grading of RCC yielded a pooled sensitivity of 0.84 (95% confidence interval [CI] 0.77-0.89) and specificity of 0.57 (95% CI 0.51-0.63) for identifying high-grade disease; the summary area under the curve (AUC) was 0.71. Low-grade tumours showed significantly higher apparent diffusion coefficient values than high-grade tumours (mean difference 0.21 × 10⁻³ mm²/s; 95% CI 0.11-0.30 × 10⁻³ mm²/s). Across five MRI-inclusive radiomics/deep learning studies, pooled sensitivity was 0.79 (95% CI 0.64-0.89) and specificity was 0.86 (95% CI 0.74-0.93), with an AUC of 0.90.
CONCLUSION: Quantitative MRI, particularly diffusion-derived metrics, shows modest accuracy for identifying WHO/ISUP grade in RCC. MRI-inclusive radiomics/deep learning models achieve higher diagnostic performance, albeit with less consistency across studies. Standardised multiparametric protocols, external validation, and decision-impact studies are required before clinical implementation.
KEY POINTS: Question Can quantitative MRI biomarkers, particularly diffusion metrics, non-invasively identify World Health Organisation/International Society of Urological Pathology grade in renal cell carcinoma? Findings Across 20 studies, diffusion showed modest grading performance, whereas MRI-inclusive radiomics/deep learning models achieved higher pooled diagnostic accuracy but remained heterogeneous. Clinical relevance Quantitative magnetic resonance imaging supports non-invasive renal cell carcinoma grading and biopsy triage; externally validated radiomics/deep learning models may improve preoperative risk stratification.
引用
- Kiani I, Fujiwara M, Mohammadzadeh S, et al. Quantitative MRI for World Health Organisation/International Society of Urological Pathology Grading of Renal Cell Carcinoma: a systematic review and diagnostic meta-analysis. Eur Radiol. 2026. PMID: 42758284. doi:10.1007/s00330-026-12837-5
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。