多参数磁共振成像影像组学在预测子宫内膜癌淋巴脉管间隙浸润中的研究进展

Research Progress of Multiparametric MRI Radiomics in Predicting Lymphovascular Space Invasion in Endometrial Cancer

摘要: 淋巴脉管间隙浸润(Lymphovascular Space Invasion, LVSI)被定义为肿瘤直接侵袭边界外的内皮细胞空间中存在肿瘤细胞，是子宫内膜癌(Endometrial Cancer, EC)重要的独立预后因素，2023年国际妇产科联盟(FIGO)新版分期系统已将LVSI纳入分层标准。然而，LVSI状态依赖术后病理诊断，术前无创评估一直是临床难题。近年来，基于多参数磁共振成像(mpMRI)的影像组学和深度学习技术，通过从医学图像中高通量提取定量特征，为术前预测LVSI开辟了新路径。本文系统梳理了该领域的研究进展：从影像组学的技术流程出发，综述了基于单一影像组学特征、临床–影像组学联合模型、深度学习模型、多区域特征融合以及生境分析等多种建模策略的研究现状与诊断性能，分析了现有研究在样本量、方法学标准化、模型泛化性及临床转化等方面的挑战，并对多模态融合与多中心前瞻性验证的未来方向进行了展望。影像组学正逐步成为EC术前精准风险分层的有力工具。

Abstract: Lymphovascular space invasion (LVSI) is defined as the presence of tumor cells within endothelial-lined spaces beyond the invasive front of the tumor, and is an important independent prognostic factor in endometrial cancer (EC). The 2023 updated FIGO staging system has incorporated LVSI into its risk stratification criteria. However, LVSI status relies on postoperative pathological diagnosis, and preoperative non-invasive assessment remains a clinical challenge. In recent years, radiomics and deep learning techniques based on multiparametric magnetic resonance imaging (mpMRI) have opened new avenues for preoperative LVSI prediction by enabling high-throughput extraction of quantitative features from medical images. This review systematically summarizes the research progress in this field: starting from the technical workflow of radiomics, it reviews the current status and diagnostic performance of various modeling strategies—including single radiomic features, clinical-radiomic combined models, deep learning models, multi-region feature fusion, and habitat analysis—and analyzes the challenges related to sample size, methodological standardization, model generalizability, and clinical translation. Finally, it offers perspectives on future directions such as multimodal fusion and multicenter prospective validation. Radiomics is gradually emerging as a powerful tool for precise preoperative risk stratification in EC.