Page 42 - 南京医科大学自然版

南京医科大学学报（自然科学版） 第45卷第11期
·1572 · Journal of Nanjing Medical University（Natural Sciences） 2025年11月

·专题研究：肿瘤·

基于影像组学的胃癌术前阳性淋巴结比率预测模型的研究

陈柃池 ，周乾正 ，李 琼 ，李沣员 ，徐 皓 1*
1
1，2
1
1
南京医科大学第一附属医院普外科，江苏 南京 210029；海安市中医院普外科，江苏 南通 226600
1 2
［摘 要］ 目的：构建并验证一个基于术前增强CT影像组学特征联合临床因素的列线图模型，用于术前预测胃癌患者的阳性
淋巴结比率（lymph node ratio，LNR），以期为个体化治疗决策提供依据。方法：回顾性收集380例行胃癌根治术患者的临床及
术前增强CT资料，按7∶3的比例划分为训练集（n=266）与验证集（n=114）。使用PyRadiomics平台提取动脉期和静脉期图像的
影像组学特征，经特征筛选后，采用支持向量机（support vector machine，SVM）、随机森林（random forest，RF）和逻辑回归（logistic
regression，LR）3种机器学习算法构建预测模型。通过受试者工作特征（receiver operating characteristic，ROC）曲线下面积（area
under the curve，AUC）、决策曲线分析（decision curve analysis，DCA）评估模型性能。将显著预测因素纳入多因素Logistic回归分
析并构建列线图。结果：RF 模型预测性能最佳，其在训练集和验证集中的AUC 值分别为0.733和0.778。多因素分析确定性
别、临床N分期（cN）和影像组学评分（Rad⁃score）是LNR的独立预测因素（P均<0.05）。基于上述因素构建的列线图模型在验
证集中表现出优异的预测效能，AUC为0.821，且DCA显示其具有较高的临床净获益。结论：成功构建了一个融合影像组学与
临床因素的列线图模型，能够在术前有效预测胃癌患者的LNR状态，有助于识别高危患者并指导个体化治疗。
［关键词］ 胃癌；淋巴结转移；阳性淋巴结比率；影像组学；机器学习；列线图
［中图分类号］ R735.2 ［文献标志码］ A ［文章编号］ 1007⁃4368（2025）11⁃1572⁃08
doi：10.7655/NYDXBNSN250829

Prediction model for positive lymph node ratio of gastric cancerbased on radiomics

1，2 1 1 1 1*
CHEN Lingchi ，ZHOU Qianzheng ，LI Qiong ，LI Fengyuan ，XU Hao
1
Department of General Surgery，the First Affiliated Hospital of Nanjing Medical University，Nanjing 210029；
2 Department of General Surgery，Hai’an Traditional Chinese Medicine Hospital，Nantong 226600，China

［Abstract］ Objective：To develop and validate a nomogram model based on preoperative contrast⁃enhanced CT radiomic features
combined with clinical factors for predicting the lymph node ratio（LNR）in gastric cancer patients，aiming to provide a basis for
individualized treatment decision ⁃ making. Methods：Clinical data and preoperative enhanced CT images of 380 patients who
underwent radical gastrectomy were retrospectively collected and divided into a training set（n=266）and a validation set（n=114）in a
7∶3 ratio. Radiomic features were extracted from arterial and venous phase images using the PyRadiomics platform. After feature
selection，three machine learning algorithms，namely support vector machine（SVM），random forest（RF），and logistic regression（LR），
were employed to build prediction models. Model performance was evaluated using the area under the receiver operating characteristic
（ROC）curve（AUC）and decision curve analysis（DCA）. Significant predictors were incorporated into multivariate logistic regression
to construct a nomogram. Results：The RFmodel demonstrated the best predictive performance，with AUC values of 0.733 and 0.778 in
the training set and validation set，respectively. Multivariate analysis identified sex，clinical N stage（cN），and radiomic score（Rad⁃
score）as independent predictors of LNR（all P < 0.05）. The nomogram incorporating these factors showed excellent predictive efficacy
in the validation set，with an AUC of 0.821，and DCA indicated favorable clinical net benefit. Conclusion：A nomogram integrating
radiomics and clinical factors was successfully developed and validated for the preoperative prediction of LNR in gastric cancer
patients，which can effectively predict the LNR status of gastric cancer patients preoperatively，helping to identify high⁃risk patients
and guide individualized treatment strategies.
［Key words］ gastric cancer；lymph node metastasis；positive lymph node ratio；radiomics；machine learning；nomogram
［J Nanjing Med Univ，2025，45（11）：1572⁃1579］

［基金项目］ 国家自然科学基金（82373335）；江苏省科教能力提升工程（江苏省医学重点学科）（ZDXK202222）
通信作者（Corresponding author），E⁃mail：hxu@njmu.edu.cn（ORCID：0000⁃0001⁃5827⁃1821）
∗