磁共振成像及影像组学在直肠癌淋巴血管浸润中的应用

Application of Magnetic Resonance Imaging and Radiomics in Lymphovascular Invasion of Rectal Cancer

摘要: 直肠癌(Rectal Cancer, RC)常见的消化系统恶性肿瘤，发病率持续上升且呈年轻化趋势，淋巴血管浸润(Lymphovascular Invasion, LVI)作为其关键不良预后因素，与肿瘤转移、复发密切相关，术前精准评估LVI对临床诊疗意义重大。当前LVI诊断依赖术后病理，存在有创、滞后等局限，而磁共振成像(Magnetic Resonance Imaging, MRI)与影像组学为无创术前预测提供了新方向。常规MRI可通过肿瘤形态学特征辅助评估，基于T₂WI的影像组学模型展现出高效预测效能。扩散加权成像(DWI)和体素内不相干运动成像(IVIM)，能从分子层面反映肿瘤异质性，与LVI状态显著相关；动态对比增强MRI (DCE-MRI)的定量参数及时间-信号强度曲线，可量化肿瘤血供与血管通透性，助力LVI评估；新兴的mDixon-Quant技术通过定量参数反映肿瘤代谢与脂质变化，具备潜在预测价值。影像组学通过从多参数MRI图像中高通量提取定量特征，结合机器学习构建预测模型，能够有效量化肿瘤异质性，提升LVI的术前无创预测准确性。基于T₂WI、DWI、IVIM、DCE-MRI等的单模态或多模态影像组学模型具有良好的诊断效能，有望无创、精准预测直肠癌LVI，为术前分期、治疗方案制定及预后评估提供重要参考，未来需扩大样本量、开展多中心研究，进一步提升模型泛化性与临床应用价值。

Abstract: Rectal cancer (RC) is one of the most common malignant tumors of the digestive system, with a continuously increasing incidence rate and a tendency toward younger onset. Lymphovascular invasion (LVI) serves as a crucial adverse prognostic factor for rectal cancer and is closely correlated with tumor metastasis and recurrence. Preoperative accurate evaluation of LVI is of great significance for clinical diagnosis and treatment. At present, the diagnosis of LVI relies on postoperative pathological examination, which is limited by invasiveness and hysteresis. Magnetic resonance imaging (MRI) and radiomics have provided a novel direction for non-invasive preoperative prediction of LVI. Conventional MRI can assist in the evaluation of LVI through tumor morphological features, and radiomics models based on T₂-weighted imaging (T₂WI) have demonstrated favorable predictive efficacy. Diffusion-weighted imaging (DWI) and intravoxel incoherent motion (IVIM) imaging can reflect tumor heterogeneity at the molecular level and are significantly correlated with LVI status. Quantitative parameters and time-signal intensity curves derived from dynamic contrast-enhanced MRI (DCE-MRI) can quantify tumor blood supply and vascular permeability, facilitating the evaluation of LVI. The emerging mDixon-Quant technique can reflect tumor metabolism and lipid changes via quantitative parameters, showing potential predictive value for LVI. Radiomics enables high-throughput extraction of quantitative features from multiparametric MRI images and constructs predictive models combined with machine learning algorithms, which can effectively quantify tumor heterogeneity and improve the accuracy of non-invasive preoperative prediction of rectal cancer LVI. Single-modal and multimodal radiomics models based on T₂WI, DWI, IVIM, DCE-MRI and other sequences possess satisfactory diagnostic performance. Such models are expected to realize non-invasive and accurate prediction of LVI in rectal cancer, providing an important reference for preoperative staging, therapeutic regimen formulation and prognostic evaluation. Future multicenter studies with enlarged sample sizes are required to further improve the generalization ability and clinical application value of the models.