肾脏乏脂肪血管平滑肌脂肪瘤与肾细胞癌的影像学鉴别诊断进展

Research Progress in Imaging Differential Diagnosis of Fat-Poor Renal Angiomyolipoma and Renal Cell Carcinoma

摘要: 随着影像检查技术的普及，肾脏肿瘤检出率显著增高，其中乏脂肪肾血管平滑肌脂肪瘤(renal angiomyolipoma, fpRAML)与肾细胞癌(Renal cell carcinoma, RCC)影像表现重叠度高，术前鉴别难度大。常规影像学检查及穿刺活检存在诊断效能有限、取样不足、亚型鉴别困难、存在并发症等缺陷，致使部分良性fpRAML患者接受不必要的手术治疗。为进一步明确两类肿瘤的无创鉴别诊断方法、优化临床诊疗策略，本文系统梳理国内外相关研究进展，围绕CT、MRI、超声影像学技术，结合血清学指标及临床基线特征，总结各类检查手段在fpRAML与RCC鉴别诊断中的应用价值与局限性。现有研究显示，CT可通过病灶形态征象、定量参数、血管分布模式、纹理分析、动态强化特征等初步鉴别；MRI凭借化学位移成像、信号衰减特征及动态强化参数，在鉴别fpRAML与RCC方面有一定的价值；超声及超声造影凭借便捷、无创优势适用于临床初筛，病灶特征性高回声征、超声造影特征具备良好鉴别价值。与此同时，基于影像组学与深度学习的人工智能分析技术也为无创诊断提供了新路径，但仍存在模型泛化能力不足等现实局限。单一影像学技术存在固有局限，而融合影像学定量指标、血清碱性磷酸酶水平及性别、年龄等临床特征的多维度联合诊断模型，可显著提升鉴别诊断准确率，临床应用前景显著。

Abstract: With the popularization of imaging techniques, the detection rate of renal tumors has increased markedly. Fat-poor renal angiomyolipoma (fpRAML) and renal cell carcinoma (RCC) exhibit highly overlapping imaging features, resulting in great difficulties in preoperative differential diagnosis. Conventional imaging examinations and percutaneous biopsy are limited by insufficient diagnostic efficiency, inadequate sampling of small lesions, difficulty in distinguishing tumor subtypes, and potential complications, which may lead to unnecessary surgical resection in patients with benign fpRAML. To further explore noninvasive differential diagnostic approaches and optimize clinical management strategies, this article systematically reviews recent domestic and international research progress. We summarize the application value and limitations of computed tomography, magnetic resonance imaging, and ultrasound, as well as serological biomarkers and clinical baseline features in differentiating fpRAML from RCC. Current studies have demonstrated that CT can provide preliminary differential diagnosis based on lesion morphological signs, quantitative parameters, vascular distribution patterns, texture features, and dynamic enhancement characteristics. MRI, characterized by chemical shift imaging, signal attenuation manifestations and dynamic enhancement parameters, has favorable value for distinguishing fpRAML from RCC. Conventional ultrasound and contrast-enhanced ultrasound are suitable for preliminary clinical screening due to their convenience and noninvasiveness, and characteristic hyperechoic signs and contrast-enhanced patterns show good differential efficacy. Meanwhile, artificial intelligence analytical techniques based on radiomics and deep learning have offered new avenues for noninvasive diagnosis, yet they still suffer from practical limitations such as insufficient model generalization ability. Single imaging modality has inherent diagnostic limitations. In contrast, the multidimensional combined model that integrates quantitative imaging parameters, serum alkaline phosphatase levels, and clinical indicators including gender and age can significantly improve diagnostic accuracy, holding great promise for prospective clinical application.