Zhang Yu, Liu Long, Han Mengyu, Li Linrui, Wu Qibing, Wang Xin
Department of Radiation Therapy, The First Affiliated Hospital of Anhui Medical University, Hefei, Anhui, China.
Department of Hepatobiliary and Pancreatic Surgery, The Second Hospital of Zhejiang University, Hangzhou, Zhejiang, China.
NPJ Precis Oncol. 2025 Jun 16;9(1):176. doi: 10.1038/s41698-025-00929-2.
Accurate assessment of lymph node metastasis (LNM) in T1/2-stage esophageal squamous cell carcinoma (ESCC) is critical for treatment planning but remains challenging due to diagnostic inaccuracies and unclear metastatic mechanisms. This study aimed to predict LNM in T1/2-stage ESCC using machine learning-based radiomics and elucidate its biological underpinnings. We retrospectively analyzed 374 surgically treated ESCC patients from two centers, employing six machine-learning algorithms to derive an optimal radiomics score. Key pathways and genes linked to LNM were investigated via bioinformatics and experimental validation. The decision tree (DT)-based radiomics model demonstrated superior predictive performance, with AUCs of 0.933 (training), 0.887 (validation), and 0.845 (test). Bioinformatics analysis implicated tumor-lymphatic invasion pathways, with EFNA1 emerging as a potential key regulator. These findings highlight the clinical utility of radiomics for LNM prediction in early-stage ESCC and provide insights into its molecular mechanisms.
准确评估T1/2期食管鳞状细胞癌（ESCC）的淋巴结转移（LNM）对治疗方案的制定至关重要，但由于诊断不准确和转移机制不明，这一过程仍具有挑战性。本研究旨在利用基于机器学习的放射组学预测T1/2期ESCC的LNM，并阐明其生物学基础。我们回顾性分析了来自两个中心的374例接受手术治疗的ESCC患者，采用六种机器学习算法得出最佳放射组学评分。通过生物信息学和实验验证研究了与LNM相关的关键通路和基因。基于决策树（DT）的放射组学模型表现出卓越的预测性能，训练集、验证集和测试集的曲线下面积（AUC）分别为0.933、0.887和0.845。生物信息学分析表明肿瘤-淋巴管浸润通路，EFNA1成为潜在的关键调节因子。这些发现凸显了放射组学在早期ESCC LNM预测中的临床应用价值，并为其分子机制提供了见解。
