定量磁共振成像方法在肝细胞癌经动脉化疗栓塞治疗反应评估与预测中的应用
本文综述定量磁共振成像(qMRI)在预测和评估肝细胞癌(HCC)患者经动脉化疗栓塞(TACE)治疗反应中的前沿应用。HCC异质性强,TACE反应个体差异大,早期识别治疗反应对优化管理至关重要。多种qMRI方法(包括肝细胞特异性对比增强MRI、扩散成像、灌注成像、MRS、BOLD-fMRI、MRE及人工智能)已报告有前景的结果,但BOLD-fMRI和MRE研究不足,AI模型附加值需更大前瞻性研究验证。
要点
- 肝细胞特异性对比增强MRI肝胆期变异系数可量化信号异质性,可能预测TACE结局。
- 扩散峰度成像在扩散成像方法中优于体素内不相干运动及扩散加权成像。
- MRS联合DWI对TACE早期评估的AUC达1.000,AI模型测试集AUC为0.690至1.000。
核心结论
- 多种qMRI方法在预测和评估HCC患者TACE治疗反应方面已报告有前景的结果,但BOLD-fMRI和MRE研究不足且缺乏关键定量参数的成熟模型,AI模型的附加值需通过更大前瞻性研究验证。
研究设计与样本
- 综述,检索范围与纳入文献原文未报告。
诊断标准与关键数值
- 肝细胞特异性对比增强MRI肝胆期变异系数可量化信号异质性,可能预测TACE结局。
- 扩散成像方法中,扩散峰度成像优于体素内不相干运动及扩散加权成像。
- 灌注成像的曲线下面积(AUC)低于扩散成像。
- MRS联合DWI对TACE早期评估的AUC为1.000。
- 结合影像组学或深度学习与临床因素的人工智能模型在测试集中的AUC范围为0.690至1.000。
局限性
- BOLD-fMRI和MRE研究不足,缺乏关键定量参数的成熟模型。
- 人工智能模型的附加值需通过更大前瞻性研究验证。
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
This article reviews the state-of-the-art applications of quantitative magnetic resonance imaging (qMRI) in predicting and evaluating response to transarterial chemoembolization (TACE) in patients with hepatocellular carcinoma (HCC). HCC is a highly heterogeneous tumor, and its response to TACE varies significantly among patients. Early identification of treatment response is critical for optimizing management. Promising results have been reported using various qMRI methods, including hepatocyte-specific contrast-enhanced MRI, diffusion imaging, perfusion imaging, magnetic resonance spectroscopy (MRS), blood oxygen level-dependent functional MRI (BOLD-fMRI), magnetic resonance elastography (MRE), and artificial intelligence (AI). The coefficient of variation in the hepatobiliary phase of hepatocyte-specific contrast-enhanced MRI, which quantifies signal heterogeneity, may predict TACE outcomes. Among diffusion imaging methods, diffusion kurtosis imaging has outperformed intravoxel incoherent motion and diffusion-weighted imaging (DWI), while perfusion imaging has shown a lower area under the curve (AUC) compared to diffusion imaging. Combining MRS with DWI has achieved an AUC of 1.000 for early assessment of TACE response. However, BOLD-fMRI and MRE remain underexplored and lack established models with key quantitative parameters. AI models incorporating radiomics or deep learning with clinical factors have shown promising AUC values ranging from 0.690 to 1.000 in test sets. However, their added value requires validation through larger prospective studies.
引用
- Zhang J, Yan C, Wang Y, et al. Quantitative Magnetic Resonance Imaging Methods for the Assessment and Prediction of Treatment Response to Transarterial Chemoembolization in Hepatocellular Carcinoma. Acad Radiol. 2025. PMID: 40068997. doi:10.1016/j.acra.2025.02.042
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。